Bundling, Tying, and Portfolio Effects: Part 2 Case …€¦ · Bundling, Tying, and Portfolio...

DTI E

CONO

MICS DTI ECONOMICS PAPER NO. 1

Bundling, Tying, and Portfolio Effects

PART 2 - CASE STUDIES

BY PROFESSOR BARRY NALEBUFFYALE UNIVERSITY AND DAVID MAJERUS CHARLES RIVER ASSOCIATES

FEBRUARY 2003

Part2 Case Studies 7/2/03 12:51 PM

Part2 Case Studies 7/2/03 12:51 PM

1

DTI ECONOMICS PAPER NO. 1

Bundling, Tying, and Portfolio Effects

PART 2 - CASE STUDIES

BY PROFESSOR BARRY NALEBUFFYALE UNIVERSITY AND DAVID MAJERUS CHARLES RIVER ASSOCIATES

FEBRUARY 2003

Part2 Case Studies 7/2/03 12:51 PM Page 1

2


3

I Introduction .........................................................................................6

II Tetra Pak International ....................................................................7

1 Background ..............................................................................................7

2 Case allegations and issues .....................................................................7

3 Market definition and dominance.............................................................8

4 European Commission findings ...............................................................8

5 Analysis of issues...................................................................................12

6 Predatory pricing ....................................................................................15

7 European Commission remedies ...........................................................16

III Tying and the HILTI case study.................................................17

1 Background ............................................................................................17

2 European Commission’s analysis of the case........................................18

3 Economic analysis ..................................................................................19

IV GE-Honeywell merger ..................................................................22

1 Introduction ............................................................................................22

2 Background ............................................................................................22

3 The case against the merger .................................................................23

4 Why bundle ............................................................................................25

5 Negotiating bundles ...............................................................................30

6 Empirical evidence of dominance and bundling.....................................33

7 The (im)practicality of bundling engines.................................................37

8 Evidence that bundling would lead to exit by rivals...............................38

9 Conclusions ............................................................................................38

Table of contents


4

V GECAS and the GE-Honeywell merger: is the Archimedean theory of leverage Kosher? ...............40

1 Introduction ............................................................................................40

2 Aircraft leasing and a history of GECAS.................................................41

3 Market shares ........................................................................................43

4 How does the leverage work? ...............................................................44

5 Would GECAS extend its policy to Honeywell?.....................................47

6 Reasons why GECAS has this policy .....................................................50

7 Empirical support....................................................................................52

8 Should we be concerned?......................................................................60

9 Conclusions ............................................................................................61

VI Independent Service Organisations v Kodak ......................63

1 Background ............................................................................................63

2 Discussion ..............................................................................................68

3 Conclusion..............................................................................................71

VII Aspen case........................................................................................72

1 Background ............................................................................................72

2 Court decisions.......................................................................................74

3 Economic analysis ..................................................................................75

VIII Guinness and Grand Metropolitan merger ..........................79

1 Background ............................................................................................79

2 European Commission decision.............................................................80

3 Analysis ..................................................................................................83

IX Interbrew and Bass merger .......................................................87

1 Background ............................................................................................87

2 Competition Commission analysis .........................................................89



5

X SMG SRH - Scottish Radio case ..............................................94

1 Background ............................................................................................94

2 OFT analysis ...........................................................................................95


XI Foreign package holidays and insurance ..............................98

1 Background ............................................................................................98

2 MMC report............................................................................................99

3 Current market conditions....................................................................103

4 Economic analysis ................................................................................105

XII BT telephone and internet bundling .....................................107

1 Background ..........................................................................................107

2 Oftel’s analysis .....................................................................................108

3 Economis analysis ................................................................................112

XIII Mobile telephones .......................................................................114

1 Background ..........................................................................................114

2 Analysis of pricing plans.......................................................................116

3 Analysis of terminating access.............................................................119

4 Conclusion............................................................................................121

XIV Jefferson Parish Hospital ..........................................................122

1 Background ..........................................................................................122

2 Case history..........................................................................................123

3 Economic analysis ................................................................................125

XV References ......................................................................................127


6

IIntroductionIn this volume we present thirteen case studies from the United Kingdom, Europe, andthe United States. These cases provide us with the opportunity to apply the issues thatare discussed in the theory volume to real world situations. For each case, we presentthe background, the decision reached on the case, and an analysis of the relevanteconomic issues. Generally we are not analysing all of the issues that were relevant tothe case but focusing on those with relevance to the issues of bundling, tying, andportfolio effects. With the exception of the GE/Honeywell case and the mobiletelephones case, we were not involved in any of these proceeding, and thus we do nothave complete information on the cases.1 At times we reach conclusions that arecounter to the ones reached by the competition authorities. We present theseconclusions not to be critical of the competition authorities but to present what webelieve would have been the appropriate methodology for analysing these particularissues. That our conclusions are different from the competition authorities could be dueto our reliance upon public information (as opposed to having the full set of information,including confidential information) or it may be due to the fact that we address onlysome of the issues in the particular case. Finally, we may differ in our conclusionsbecause of different interpretations of the relevant economic theory.

1 Our analysis of the cases we were involved in, GE/Honeywell and mobile telephones, do not rely on confidential information.


7

IITetra Pak InternationalThere have been a number of significant cases involving Tetra Pak over the years,including the recent Court of First Instance overturning the European Commissionsdecision on the Tetra Pak/Sidel proposed merger. In this chapter, we review theallegations of Tetra Pak’s abuse of a dominant position in 1983. This alleged behaviourilluminates a remarkable number of issues we have discussed in this report. Accordingto the complaint, Tetra Pak was using tying and bundling to engage in everything frommetering to creating entry barriers. Thus, we are able to apply several of the theoriesdeveloped in Chapter 6 of the main report to analyse Tetra Pak’s incentives and the likelyeffects of its behaviour. We begin with a description of the company and the allegedabuses as well as the findings of the European Commission. Then we explore theincentives of the firm and the likely effect of the conduct.

1 Background

The Tetra Pak group of companies (Tetra Pak) is one of the world leaders in the field ofthe packaging of liquid and semi-liquid foods in cartons. It introduced its first packagingsystem for milk in 1952. This system included a tetrahedron shaped container, a fillingmachine, and the technology for coating the packaging material and sealing the materialafter it is filled. The Tetra Brick packaging system was introduced in 1963. The packagingtechnologies have been improved over the years. For example, Tetra Pak developed amethod for the aseptic packaging of milk, allowing for longer shelf life without the needfor refrigeration. Through a number of acquisitions, Tetra Pak has expanded into otherforms of packaging, including plastic packaging. Currently Tetra Pak operates in both thefield of the non-aseptic packaging of fresh products and in aseptic packaging of long-lifeproducts. Tetra Pak’s consolidated turnover has steadily grown over time, from C2 billionin 1985, to C2.4 billion in 1987 and approximately C3.6 billion in 1990.

Tetra Pak has facilities across Europe and around the world. Tetra Pak’s first commercialoperations in Italy - the country at the heart of the alleged abuses - date back to 1955. In 1965, the first carton production unit was opened and in 1980 a machine assemblyplant was started up. The consolidated turnover of the seven Italian companies withinTetra Pak stood at C204 million in 1987.

2 Case allegations and issues

Elopak Italia filed a complaint in 1983 claiming that Tetra Pak had engaged in a number ofpractices that were abusing its dominant position. Elopak was Tetra Pak’s maincompetitor in the field of non-aseptic packaging of liquid foods in cartons. These abuses


8

broadly fell into three different areas: selling cartons at predatory prices, imposing anti-competitive conditions on the supply of machines for filling these cartons, and selling the filling equipment at predatory prices. These alleged anti-competitive practices alsoincluded alleged attempts at excluding Elopak from certain advertising media. The Commission investigation was initiated after the filing of this complaint.

PRODUCTS

Tetra Pak’s packaging systems all rely on some type of carton for holding the packagedproduct. Although the carton is well suited to the packaging of a wide variety of liquidand semi-liquid foods, figures from 1987 show the use of this packaging was primarilyused in the dairy industry: 79% of cartons were used for packaging milk (72%) and otherliquid milk products (7%), 16% for fruit juice, and other food products togetheraccounted for only approximately 5% of cartons used. When the original complaint wasfiled in 1983, some 90% of cartons were used for the packaging of milk or milkproducts. At the time of the inquiry, 60% of milk within the Community was sold incartons and 50% of fruit juice consumed in the Community was packaged in cartons.

MILK PACKAGING

Milk is sold in a pasteurised form (fresh milk) or after ultra-high temperature treatmentunder aseptic conditions (UHT milk), the latter process allows the milk to be stored forup to several months in a non-refrigerated environment. The milk-processing techniques,including the packaging process are different for these two forms of processed milk. In particular, the UHT process has more demanding pressure and sterilisation requirements.Therefore, the packaging machines and cartons for these two processes are notinterchangeable. It is unlikely to have much substitution between the two different typesof machines (and cartons) even if there was a sustained price increase of one of thepackaging systems. This lack of substitution is driven by several factors: the cost of thepackaging was small relative to the total cost of the final products, consumers havedifferent perceptions of the types of packaging (consumers would not expect to find freshmilk in a Tetra Brick container), and the filling lines were not able to switch between typesof packaging without significant investments. All of these factors lead to the conclusionthat aseptic packaging systems are in a separate relevant market than non-aseptic systems.

3 Market definition and dominance

As is often the case, there was a difference of opinion as to the correct relevant marketdefinitions. The commission concluded that there were four separate markets:

(i) the market for machinery incorporating technology for the sterilization of cartons andthe packaging in those cartons, under aseptic conditions, of UHT-treated liquid foods;

(ii) the corresponding market for aseptic packaging cartons;

(iii)the distinct (but neighbouring) market for machinery enabling fresh liquid foods to bepackaged in cartons; and

(iv) the corresponding market for non-aseptic packaging cartons.


9

Tetra Pak argued that the relevant market was the overall liquid food packaging market,encompassing all forms of packaging.

In examining the substitutability of the products involved, the Commission came to theconclusion that while different forms of packaging as diverse as glass bottles, plasticbottles, plastic bags, metal tins, aseptic cartons and non-aseptic cartons formed part ofwhat was commonly known as the packaging market for liquid foods, this was not therelevant market within the meaning of Article 86. These different types of packagingcompeted with each other only in the long term. “In the short and probably evenmedium term, the conditions of supply and demand were such that the elasticity ofsubstitution for products in relation to prices was almost zero.”1 The Commission alsofound that the relevant geographical market was the entire European Community.

Tetra Pak’s share of the aseptic packaging and filling machine markets was found to bebetween 90 and 95 percent. Given its high share and the existence of high barriers toentry for the packaging machines, the Commission ruled that Tetra Pak was dominant inthe aseptic packaging markets. In the non-aseptic packaging, Tetra Pak had a share ofaround 45-50%. The Commission did not reach a conclusion as to whether this positionrepresented one of dominance. The Commission did not consider it necessary toaddress this issue, given the association between the non-aseptic packaging marketsand the aseptic packaging markets.2 In particular, the Commission found that “Tetra Pakhas used the association which exists between the four markets in question to commitabuses on the non-aseptic markets, abuses which it could not have committed in theabsence of its dominant position on the aseptic markets.”3

4 European Commission findings

Tetra Pak’s share in the aseptic packaging market made it essentially the only supplier forfirms that produced aseptically packaged liquid foods. Tetra Pak allegedly used itsposition in the aseptic packaging market to force its customers to accept contractualconditions that prevented any inter- or intra-brand competition. In what follows, wemention only the main contractual conditions that were judged abusive under Article 86.

SALES CONTRACTS

The first series of abuses arose out of the sales contracts used by Tetra Pak when sellingthe filling machines. These sales contracts prohibited customers from modifying, addingaccessory equipment to or moving (and thus exporting) machines. Other features of the

1 Commission decision 92/163/EEC of 24 July 1991 relating to a proceeding pursuant to Article 86 of the EEC Treaty (IV/31043 - TetraPak II) at ¶ 93.

2 Tetra Pak appealed the decision to the Court of First Instance in part based on the Commission not determining whether or not TetraPak was dominant in the non-aseptic packaging markets. The CFI confirmed the Commission's position, thereby acknowledging thatabuses could be committed in markets in which a firm was not dominant, provided there were close links to the market(s) in whichthat firm was dominant.

3 Commission decision 92/163/EEC of 24 July 1991 relating to a proceeding pursuant to Article 86 of the EEC Treaty (IV/31043 - TetraPak II) at ¶ 104.


10

contracts included the obligation to obtain Tetra Pak’s agreement before selling usedequipment; priority right of repurchase by Tetra Pak at a prearranged price; the obligationto ensure that any third party acquiring the equipment assume all of the initial buyer’sobligations vis-à-vis Tetra Pak; the obligation to inform Tetra Pak of any improvements ortechnical modifications to equipment or to cartons and to grant Tetra Pak ownership ofany resulting intellectual property right; and, finally, the right for Tetra Pak to inspect thewording to be used on cartons. All maintenance and repairs on the machines were to becarried out by Tetra Pak. Tetra Pak also secured the exclusive right to supply spare partsfor the machines and adopted a sliding scale for charges made for assistance andmaintenance, depending on the number of cartons used on Tetra Pak machines.

In addition, contractual conditions required that only Tetra Pak cartons be used on theseTetra Pak machines. Also there was the requirement that Tetra Pak cartons be acquiredfrom the group’s local subsidiary of Tetra Pak (i.e. no independent distributors). Thewarrantee of the machines was conditional on compliance with all contractual conditions(in certain contracts) or, at the very least, exclusive use of Tetra Pak cartons.

The customers of the filling machines had to provide a monthly report on carton use and had to allow Tetra Pak to inspect the machines. These inspections gave Tetra Pakthe right to inspect, without notice, production sites, accounting or other documents and customers’ correspondence.

Tetra Pak offered customers the choice of buying or leasing the machines. Equipmentleases included the same restrictive provisions as sales contracts. Moreover, thestructure of rent payments made leasing indistinguishable from buying, since the rentalcontract included, upon delivery, an initial payment corresponding to virtually all presentand future rental charges (more than 98 per cent in some cases).

These contractual terms were justified based on three reasons: technical grounds,liability considerations with regard to public health, and that such requirements werestandard in other commercial situations. Without denying the complexity involved, theCommission rejected Tetra Pak’s first two arguments, saying that to the extent therewere technological and liability issues, they could be solved through means that wereless restrictive than the existing contractual requirements. With regard to commercialusage, the Commission disputed that such practices were common in the non-asepticpackaging markets and even if they were common in a competitive market would notjustify them in the aseptic market where Tetra Pak had a dominant position.

PREDATORY PRICES

The combination of its high share, barriers to entry, and the contractual restrictions listedabove resulted in Tetra Pak being able to reap the benefits of its significant marketpower in the aseptic packaging markets. The Commission found that the extremeprofitability of aseptic packaging, which contributed more than 90 per cent of corporateprofits, provided Tetra Pak the ability to make financial sacrifices in the non-asepticmarkets without in any way jeopardising overall margins. A general analysis of the


11

profitability of the various Tetra Pak products, carried out using the group’s own costaccounting, showed that in some Member States “aseptic bricks” - the best selling andmost profitable cartons had been subsidising loss-making sales of “non-aseptic bricks”or even all of Tetra Pak’s other products.

The Commission concluded that the sales of the non-aseptic cartons at below cost werenot from exceptional circumstances, but rather that they had been the result of adeliberate policy calculated to eliminate the competition. The Commission described theexistence of predatory pricing over a period of time for machines in the United Kingdomand for cartons and machines in Italy.

The Commission found that the predatory pricing practices were especially serious inItaly, claiming that but for the Commission’s intervention, these practices probably wouldhave led to the total elimination of Elopak, Tetra Pak’s chief rival and leader in thatcountry’s non-aseptic packaging market. Apart from many occasions in which machineswere sold at a loss, for seven years Tetra Pak had sold its Rex cartons in Italy - Rex beinga competitor of Elopak’s main product, the Pure-Pak carton - at prices well under itsdirect variable costs. The 30-35 per cent difference between the price of a Rex cartonand its direct variable costs was almost exactly equal to the difference between theprices at which Tetra Pak’s Rex and Elopak’s Pure-Pak cartons were being sold.

REMEDIES

In its decision, the Commission ordered Tetra Pak to cease all of its abusive behaviourand to take a series of measures to open up its markets. In particular, Article 3 of thedecision ordered that Tetra Pak take the following measures:

(i) amend or delete the clauses from its machine purchase/lease contracts and cartonsupply contracts so as to eliminate the abuses noted above;

(ii) practice neither predatory nor discriminatory pricing and refrain from granting anycustomer any form of discount on its products or more favourable payment terms notjustified by an objective consideration. Thus, discounts on cartons were to be grantedsolely according to the quantity of each order, and orders for different types of cartonswere not to be aggregated for that purpose;

(iii)refrain from refusing orders, at prevailing prices, even if the company placing theorder was not an end-user of Tetra Pak products; and

(iv)inform any customer purchasing or leasing a machine of the specifications packagingcartons had to meet in order to be used on that machine.

Shortly before the Commission’s decision, Tetra Pak sent the Commission a draft reformof its policy and contracts, meeting some of the Commission’s demands. After thedecision, and despite having lodged an appeal with the judicial authorities of the EuropeanUnion, the group stated that it was ready to comply with the Commission’s decision. It amended its contracts accordingly, and initial reports on the application of the decision(which the Commission requires for five years) have been positive. Based on the information


12

conveyed in the first report, Tetra Pak is believed to have ceased its discriminatorypricing practices following the Commission’s decision. In addition, a C75 million fine wasimposed on Tetra Pak - at that time the largest fine ever levied on a single firm.

5 Analysis of issues

Tetra Pak is a remarkable case in that the cited behaviour includes almost a full catalogueof the potential anticompetitive activities associated with bundling and tying. The citedbehaviour included metering for price discrimination, elimination of complement markets,and elimination of potential entrants through bundling.

METERING OF ASEPTIC FILLING MACHINES

A firm with market power will want to price discriminate among consumers. Metering isoften the route to price discrimination. To be successful in metering, a firm must find agood used by customers in proportion to the customer’s value. In addition, the firmneeds to make it difficult for its customers to avoid the metering product. As in the HILTIcase discussed in Chapter III below, the ideal meter is a per-use fee. Tetra Pakeffectively achieved this result through a combination of the various restrictive clauses inthe contracts for its filling machines along with the requirement that only Tetra Pakcartons be used. Although one does not want to applaud potentially illegal behaviour, weare impressed, perhaps even taken aback, by the extent to which Tetra Pak was able toput together all of the ingredients for a successful metering strategy.

As explained in Section 6.6 of the main report, when a firm tries to price discriminate,consumers will incur costs as they try to disguise their value. It appears that Tetra Pakmade it quite difficult, if not impossible, for consumers to disguise their types. Tetra Pakrequired that its customers use only Tetra Pak cartons on its machines, so the cartonwas the metering device. To discourage use of unmetered cartons, Tetra Pak (as part ofthe machine contract) had the right to inspect the machines. These inspections wouldallow Tetra Pak to monitor the type of carton being used. It also would help Tetra Pakestimate the size of the firm, to see if the number of cartons being purchased from TetraPak were the correct order of magnitude. By performing all of the maintenance on themachines, Tetra Pak could also estimate the amount of use of the machines to ensurethat an appropriate number of cartons were purchased. Finally, by not allowingcustomers to resell the machines (without Tetra Pak’s permissions), Tetra Pak was ableto assure that even purchasers of used machines were subject to its requirements.

Given that tying the cartons to the machines is a near perfect metering device, therewas likely little direct loss of social welfare from these actions.4 If all of the customersvaluations were the same and varied only according to their use of cartons then TetraPak would be able to extract all (or nearly all) of the consumer surplus from the users ofits aseptic packaging machines. But, in reality, there were different valuations betweenconsumers. Thus Tetra Pak needed an additional device for price discrimination. For example, there could be differences across countries as to the level of demand for


13

UHT milk. Thus charging a single price for all cartons across Europe would have led toimperfect price discrimination.

To get around this problem, Tetra Pak charged different prices for cartons in differentcountries and required users to purchase cartons only from the local authorised dealer.Tetra Pak adopted a market compartmentalisation policy, which allowed for persistent pricedifferences for cartons among different countries.5 Table 1 provides the prices charged in different countries for 1981-4. Table 2 provides similar information for earlier years.

While there might not be much loss of efficiency from this pricing behaviour, it does leadto a significant transfer of surplus from consumers to Tetra Pak.

4 There could be a loss of economic efficiency associated with Tetra Pak expending resources in an effort to monitor its customer’susage of cartons. For example, if Tetra Pak has to inspect the machines regularly, the costs of these inspections will be wastedresources. A second loss would be customers that switched to other, less preferred, packaging solutions in order to maintain someconsumer surplus. But, as Tetra Pak was able to charge different prices to different customer types, they should have been able tominimise this loss.

5 The distribution of cartons and machines was undertaken directly by companies within the Tetra Pak group in each state member ofthe European Community. There were no independent distributors.

TABLE 1

AVERAGE PRICE OF CARTONS - INDEX: ITALY=100

Carton Year Italy Netherlands Germany France UK Denmark Belgium

Rex 1981 100 129 117 124 1341981 100 151 136 122 1461983 100 146 140 117 138 1481984 100 127 124 106 133 136

Non- 1981 100 102 122 105 128aseptic 1981 100 105 136 124 149brick 1983 100 97 151 116 151

1984 100 78 129 105 133

Aseptic 1981 100 156 133 167 114 139brick 1981 100 170 143 163 119 139

1983 100 162 142 142 113 1401984 100 137 123 127 109 120

Source: Commission Decision 92/163/EEC Of 24 July 1991, Annex V.


14

ENTRY BARRIERS

The biggest threat to Tetra Pak’s market power would appear to be new entry6. Severalof Tetra Pak’s policies seemed designed to solve this problem.

One entry deterrence strategy was to limit the size of the market that would be available to a new entrant at any point in time. Because the purchase (or lease) terms for themachines had different expiration dates across the different consuming firms, Tetra Pakwas able to make entry more difficult. At any point in time, only a fraction of the users of aseptic packaging were available to a potential new entrant, making entry less likely to be profitable than if the entrant were able to compete for the entire market. By havinga clause that sets the price at which Tetra Pak can repurchase the machines from itscustomers, this also could make it more costly for a firm to switch to an alternativeprovider.7 These practices helped Tetra Pak protect its market from entry.

Tetra Pak further discouraged entry by its control over the complementary repair and spareparts markets (see also the Kodak case in Chapter VI below for comparison). Tetra Pakhad at least two possible reasons for imposing exclusivity deals on maintenance and the provision of parts. First, it allowed Tetra Pak to monitor the use of the machines to

6 Tetra Pak’s concern about entry was demonstrated by its reaction to the attempted entry of an aseptic packaging system into Italy.This reaction is documented in ¶¶ 76 to 83 of Commission decision 92/163/EEC of 24 July 1991 relating to a proceeding pursuant toArticle 86 of the EEC Treaty (IV/31043 - Tetra Pak II).

7 We are not aware of the exact structure of these purchase-back prices, but presumably Tetra Pak could set these to be at a low pricefor a given term, and then rising beyond that. Such a pricing structure would encourage the consuming firm to keep using the TetraPak machine until the repurchase price rises to a higher level. By staggering these trigger dates, Tetra Pak could in theory limit thefraction of the market that would be available to a new entrant at any given point in time.

TABLE 2

PRICE OF REX CARTONS ACCORDING TO PRICE LISTS - INDEX: ITALY=1001

State Year

1978 1979 1980 1981 1982 1983 1984

Italy2 100 100 100 100 100 100 100Germany3 152 171-192 180 121Denmark 141 140 132 168 164 150 127United Kingdom 172 148 138 131France4 150 125Belgium 144 125Netherlands 126 125 120 130-165 158 139 115

Notes.

1 Indices are calculated on the basis of prices converted into C. One litre Rex cartons, price per order of 200,000 units.

2 Minimum orders of 250,000 units.

3 Minimum orders of 500,000 units.

4 Minimum orders of 1,000,000 units.

Source: Commission decision 92/163/EEC of 24 July 1991, Annex V.


15

ensure that all of the conditions of the contract were being adhered to. Second, itprevented the creation of an industry that could provide maintenance and parts for thefilling machines. If such an industry existed, it would lower the costs of entry by allowingthe entrant to take advantage of the competitive maintenance sector.

Similarly, the requirement to use only the Tetra Pak cartons on the filling machines waslikely imposed for two different reasons. First, the carton was a metering device andallowed Tetra Pak to extract more surplus out of the high-value users. Second, Tetra Pakwas able to keep new entrants out of the carton market by limiting the size of themarket for cartons. Since entrants would have available to them only the part of themarket that uses other companies’ filling machines, this would be a small market.8

To be successful, entry would have to be at both the carton level and the machine level -a harder task than just at the carton level. While any of these actions might have alimited effect on raising entry barriers, all of them together could have a significant effectin decreasing the probability of entry.

6 Predatory pricing

In its decision, the Commission claims to have demonstrated that Tetra Pak was pricingbelow its costs for some non-aseptic cartons. In its decision, the Commission suggeststhat the predatory pricing was financed by the profitability of the aseptic market, but wedo not find this last argument persuasive. We believe that there could well have beenpredation. What we do not find persuasive is that the profitability of the adjacent marketis a relevant factor in analysing whether or not there is predation. The Commission madethe argument that Tetra Pak’s significant profits in the aseptic packaging market allowedit to predate. We think the analysis of predation should focus on whether the firmexpects the predation strategy to be profitable. This is the general stumbling block topredation. If predation is expected to be profitable, then a firm should be able to financethis predation either internally or through external capital markets.

An interesting aspect of Tetra Pak’s predation is that it could be following this strategyfor one of two reasons. First, Tetra Pak can be selling non-aseptic cartons below costwith an eye towards eventually monopolising this market. For this to be a rationalstrategy, Tetra Pak must be able to raise prices on the non-aseptic products in the future.The other reason for below-cost pricing is to raise entry barriers in the aseptic market. In other words, by making entry less profitable in an adjacent market (the non-aseptic),entry is harder into the aseptic market. Which of these two is taking place can bedetermined from the way that Tetra Pak is pricing the below cost products. If these lowprices were only available as part of a bundle of products (i.e. if the customer buysaseptic products then it gets a lower price on the non-aseptic products), then thepredation is likely part of an entry deterrence strategy. On the other hand, if the belowcost prices are available to all customers, regardless of other purchases, then it is likely to be a case of old-fashioned predation. In this later case, the predation has nothing to dowith Tetra Pak’s market power in aseptic packaging.

8 Limiting the size of the carton market available to a new entrant only matters if there are economies of scale in producing cartons.


16

We do not have direct evidence, from the Commission report, of direct discountingacross different products. However, we do know that roughly 60% of the users ofproducts offered by Tetra Pak in the non-aseptic market were also active in aseptic. In terms of volumes, the percentage is higher due to the presence of some big dairiesand bottlers of fruit juices. We also learn from the report that in Italy, Elopak’s fall in sales were concentrated in those dairies that produced both fresh and UHT milk (whichimplies that they were customers of Tetra Pak in the aseptic market). Finally, one of theremedies imposed by the commission explicitly addressed the issue of discountsgranted on orders of different types of cartons, by forbidding it to aggregate orders ofdifferent types of cartons to calculate quantity-based discounts. Thus, we suspect thatTetra Pak’s strategy included a bundled pricing strategy that resulted in below cost salesof some machines as an entry deterrence tool.

7 European Commission remedies

The Commission’s remedies appear to relieve the competitive concerns mentionedabove. While we are not aware of the form of the contract that Tetra Pak now has withits customers for the filling machines, the concerns mentioned by the Commission in its decision suggest that it will ensure that offending clauses are not included in thecontract. In particular, the Commission will not allow Tetra Pak to require that itscustomers only use its cartons on its filling machines. This will ensure that Tetra Pak isnot able to meter its customers (at least not in the same manner it was using). Further,the Commission required that discounts be based solely on the purchases of the productbeing discounted, so there can be no prices that discount one product based on thepurchases of another product. This remedy should ensure that Tetra Pak is not able touse a bundling strategy to disadvantage firms in only one industry. Finally, theCommission has required that Tetra Pak not charge “uneconomic prices” in an effort tomake sure there is no predatory pricing in the future.

While we have seen cases where a particular bundling or tying activity can be givenmultiple interpretations, this Tetra Pak case is impressive for its ability to demonstrate somany potentially anticompetitive practices all at the same time. Tetra Pak in 1983 wouldappear to be the poster child for anticompetitive bundling, tying, and portfolio effects.


17

IIITying and the HILTI case studyThis case provides an example of a tying strategy that is a puzzle according to theChicago School theory. As elaborated in section 3.2 of the main report, there is noeconomic incentive to tie when there are fixed proportions between the tied and tyinggood, as was the case with HILTI’s cartridges and nails. As a result, we are inclined toaccept HILTI’s argument that the bundling was done for reasons of quality and safety.1

1 Background2

In 1986, HILTI Aktiengesellshaft (HILTI AG) was a world leader in the manufacture anddistribution of fastening systems (nail guns, drilling equipment, nails, cartridges, andcartridge strips) for use in the building industry. Based in Liechtenstein, HILTImanufactured products in Germany and elsewhere in the EEC. In 1986, HILTI hadworldwide turnover of almost 1.5 billion Swiss Francs.

Nail guns first became popular in 1958, when Dr. Martin HILTI perfected the first version.The nail guns at the heart of this case operate using an exploding cartridge that propels a piston to drive in the nail (or other type of fastener). The nail and cartridge are separateparts although they can be sold as a bundled product. Manufacturers sell a variety of gunsfor different types of fastening situations, and the strength that the particular fastener ispropelled can be adjusted by means of cartridges or power regulation systems. Poweractuated fastening systems (PAFS) encompasses nails, nail guns, and cartridge strips.

Standard nails cannot be used in nail guns, and the specially manufactured nails forthese nail guns must be adapted to fit specific guns. To some extent the different brandsof nails fit more than one brand of nail gun, so there is some interchangeability betweenthe different nails. Modern nail guns permit the user to fire multiple nails withoutreloading the gun through the use of a disk or strip containing several (usually 10)cartridges. These cartridge strips generally are made to fit specific brands of nail gunsand thus are not generally interchangeable.

The cartridge explodes as part of the use. This explosion is what provides the force to drivethe nail into the concrete or other material. As a result, each of the ten cartridges on a stripis a one-use item. If there was ever a case of a fixed proportion product, this is it. A ten-

1 In this analysis we are focussing solely on the tying claims and are not reviewing other aspects of this case.

2 The information contained in this case is obtained from the European Commission Decision of 22 December 1987 relating to aproceeding under Article 86 of the EEC Treaty (IV/30.787 and 31.488 — Eurofix-Bauco v. HILTI) as well as the Judgment of the Courtof First Instance (Second Chamber) of 12 December 1991 (HILTI AG v Commission of the European Communities).


18

pack cartridge requires precisely ten nails. The customer requires both a cartridge and anail to use the gun. Neither the nail nor the cartridge is reusable.

HILTI held patents on certain features of its nail guns. HILTI had not taken any legalaction against other companies on the basis of nail gun patent protection as of 1987. Inaddition, HILTI held patents on the cartridge strips. In the EEU, HILTI did take actionagainst a company for the sale of cartridge strips, which allegedly violated some or all ofthese patents. This competitor’s cartridge strip was ultimately removed from themarket.3 The cartridges themselves were not patented.

During the early 1980’s, Eurofix was a small company engaged in the manufacture anddistribution of nails, including a style for use in HILTI nail guns. Another small company,Bauco, specialised in the importation of nails for use in nail guns made by HILTI.

Eurofix filed a formal complaint on October 7, 1982, accusing HILTI of breaching Article86 of the EEC Treaty.4 Eurofix alleged that acting through its subsidiaries, HILTI was“pursuing a commercial strategy designed to exclude Eurofix from the market for nailscompatible with HILTI products,” and specifically that:

“HILTI refused to supply independent dealers or distributors of HILTI products withcartridge strips without a requisite complement of nails; that in response, in order tosell its nails for HILTI nail guns, Eurofix tried to obtain supplies of cartridge strips itself;that HILTI induced its independent dealer in the Netherlands to cut off supplies ofcartridge strips that Eurofix had previously obtained from this source; and that Eurofixwas also refused supplies of cartridge strips following a direct request to HILTI.”5

Bauco made a similar complaint to the Commission.

The independent nail producers (Eurofix and Bauco) thus alleged that HILTI was tying thesale of nails to the sale of cartridge strips and that this tie was resulting in significantanticompetitive effects.

The European Commission ultimately found that HILTI had violated Article 82 (then 86) ofthe EEC Treaty. As a result of this violation, HILTI was fined and also required to endthose acts which the Commission found to cause competition problems. HILTI appealedthis decision to the Registry of the Court of Justice in 1988.

2 European Commission’s analysis of the case

The Commission found that there were three separate relevant markets: a market fornail guns, a market for HILTI-compatible nails, and a market for HILTI-compatiblecartridges. Further the Commission found that HILTI was dominant in the market for nailguns. The Commission found that there was not a market for PAFS and the Commissionfurther found that the PAFS products were not part of a larger fastening systems marketas alleged by HILTI.

3 We have been told that there were some questions as to the validity of this patent on the cartridge strip. We return to this questionand the implications for the analysis below.

4 Since the time of the filing of this case, Article 86 has become Article 82.

5 European Commission Decision, 22 December 1987, IV/30.787 and 31.488 - Eurofix-Bauco v. HILTI at 25.


19

The Commission reached its conclusion of dominance in the market for nail guns in partthrough the assessment of market share estimates for nail guns provided by HILTI for1982. According to these calculations, the only competitors within the EEC with significantmarket shares were Spit and Impex. Within this market, HILTI was found to have by farthe largest share of sales.6 In addition to having dominance in nail guns, HILTI was foundto be dominant in the cartridge strips market, primarily as a result of having patents.

Given the large number of HILTI nail guns in the market, many independent nail makerswere interested in supplying nails for these guns. There were more independent nailmakers than cartridge makers. The Commission determined that HILTI’s share of themarket for HILTI-compatible nails and HILTI-compatible cartridge strips were both verysubstantial, and in excess of its share of the market for general nails and cartridge stripsrespectively.

The Commission concluded that HILTI did tie the sale of nails to the sale of cartridges in that for a number of customers HILTI refused to sell cartridges unless there weresufficient nails sold for the number of cartridges. Further, the Commission found that in those cases where there was not an absolute tie, that HILTI reduced the discount on the cartridges when the customer did not also purchase HILTI nails.

3 Economic analysis

Since we were not involved in the analysis of the case and have not had access to anyconfidential documents in this case, we will take the Commission’s analysis and suggestalternative conclusions to those reached by the Commission. Also we explore thepossible reasons, assuming the Commission’s analysis was correct, that HILTI may havehad for tying its nails to its cartridge strips.

We begin from the assumption that HILTI had an absolute tie between its cartridge stripsand the nails. The first question we address is: under what circumstances does this tielead to anticompetitive effects? It is our understanding that HILTI justified the tiebecause of safety concerns of improper nails being used in the nail guns, and suchimproper nails could be ineffective fasteners and also cause harm to the operator. Beforeaccepting this explanation, we first want to explore if there could be a strategic reasonfor this tying.

We will assume that the HILTI cartridge strip is a correctly defined antitrust market.7

Further, we will assume that HILTI has a dominant position in this cartridge market. Finally, we will assume that HILTI has this dominant position because of its patent of the cartridge system and that this patent was lawfully obtained.8 Given the nature of patents, this allows HILTI to price its cartridges at the monopoly level.

6 Specific share and sales information are not available in the publicly available documents for this case.

7 As in all antitrust analysis, the market definition is dependent upon a number of specific facts and we do not have access to thenecessary facts. It might seem reasonable that a relevant market would be “fastener systems” and that competition is at the nailgun/cartridge/nail level as opposed to just the cartridge level. This is the type of question that often arises when a durable good is soldand there is some type of consumable that is then used with the durable good.

8 As mentioned above, there was some question about the legality of this patent. We return to this question below.


20

Given these assumptions, it does not appear that there can be consumer harm from thealleged tie. HILTI has no ability to gain additional profits by pricing the nails high and tyingthem to a low-priced cartridge or pricing the nails low and the cartridge high. Since thereis a fixed number of nails that are used with each cartridge, there is no ability to meterusage through this (or any other) pricing strategy. Furthermore, we find it unlikely thatHILTI will be able to deny scale economies to other nail producers by foreclosing theHILTI nail gun nail market to them. Since there are many other uses for nails besidesHILTI nail guns, and nail manufacturing does not seem to us to be an industry withsignificant economies of scale, so we do not believe that there is the ability to driveother nail manufacturers out of the market. Therefore, on the face of it, while HILTI wastying its nails to its cartridge, we do not see any competitive harm from this behaviour.

This conclusion does raise the question of why was HILTI tying nails instead of justpricing the cartridge high and gaining its profits that way? A similar example wasdiscussed in section 3.2 of the main report, namely tying bottled water to trips on theLondon Eye. HILTI could charge up to the monopoly price for the cartridge plus nails andhow it splits this between the nail price and the cartridge price does not affectconsumers. If the competition authorities allow a high price on a patented product buttying is not allowed, then HILTI should be happy to earn all of its profits on the cartridgeand price the nails at the marginal cost of making nails.

We posed this question to an executive at HILTI. It was brought to our attention that atthe time the Commission was investigating this tying behaviour, there was also aninvestigation into the validity of HILTI’s patent on its cartridge strip. If there was a riskthat this patent would be invalidated, then we have an explanation for why HILTI wouldnot charge a high price for its cartridge strip and a low price for the nails.9 Even with aquestionable patent there still remains a mystery of why HILTI would tie its nails with itscartridge strip. While we do not know all the facts about the nail industry, we suspectthat it is reasonably competitive. As shown in section 3.2 of the main report, under thisassumption bundling the cartridge strip and nails is neither an effective entry deterrencedevice nor a way to avoid double marginalisation. The fixed proportion usage eliminatesthe metering motivation. We do not believe that this tie would eliminate the competitivecomplements market in nails (the elimination of a complementary market is discussed inthe Kodak case in Chapter VI below). So, again, we are in the situation that HILTI wouldnot be able to gain monopoly profits (since by assumption its patent is not valid), butwhatever profits it can make should be invariant to how it prices.

Having rejected all other explanations, we believe that a reasonable conclusion is thatHILTI had a motive other than increasing its profits when it tied nails to its cartridge strip.During the investigation, HILTI argued that the combination of safety and reputation weredriving its tying of nails to cartridge strips. HILTI claimed that other companies’ nails hadhigher failure rates than its own. We are not in a position to evaluate this claim, but if

9 There may have been some constraint on the price it could charge for the cartridge if the court required it to license the cartridge at aspecified price. But until such licensing was required, there was no reason not to make all of the profits on the cartridges and not onthe nails.


21

consumers would blame HILTI for the failure of a nail whether it was a HILTI nail orsome other nail used with a HILTI nail gun, then this would provide a reason to tie nailsto cartridge strips. If there is no obvious profit gain from a company tying two productsthen this provides support for a claim of quality being the motivation for tying.

In conclusion, HILTI is an example of tying that fits into the Chicago school model of firmbehaviour. There does not seem to be any ability for HILTI to gain any profits from tying(regardless of whether its patent is valid or not). In these circumstances, we believe thatthe firm should be given the benefit of the doubt in terms of its quality and safety claims for the tie.


22

IVGE-Honeywell merger1 Introduction

The economic theory of bundling has moved from the classroom and academic journalsto the public policy arena. Its debut was dramatic. On July 3, 2001, the EuropeanCommission blocked the proposed $42 billion merger between General Electric (GE) and Honeywell.1 A primary reason for their objection to this combination was a concernover bundling.

This case study uses the context of the proposed GE-Honeywell merger to address theconcerns raised by bundling.2 We set out the theory as put forth by the Commission andtry to reconcile this theory with both the economic theory of bundling and the facts ofthe case. We discuss what is meant by bundling and explain when it is a potentialproblem and when it is not. Based on this understanding, we propose antitrust policyrecommendations to deal with the novel issues raised by bundling.

2 Background

On October 19, 2000, United Technologies Corporation (UTC) reported that it was inmerger discussions with Honeywell. Three days later, a merger was announced - but thebuyer was GE, not UTC.

This case generated a good deal of attention. General Electric is one of the most well-known and admired companies in the world. At $42 billion, this was a large merger evenfor GE. The proposed integration passed the scrutiny of the US Department of Justice.Because of the size of GE and Honeywell’s European sales, the merger also had to beapproved by the European Commission.3 On July 3rd, 2001, that permission wasdenied.4 The divergence of outcomes between the US and European antitrust authoritiesadded to the publicity of this case.

1 The merger was voted on by the 20-member European Commission. Their vote confirmed the recommendation made by CompetitionCommissioner Mario Monti, who, in turn, was given a recommendation by the European Union Merger Task Force.

2 The author was an economic expert for GE-Honeywell in their presentation to the European Union Merger Task Force. The applicationof bundling theory to the GE-Honeywell merger was done together with Patrick Rey, Carl Shapiro, Shihua Lu, and Greg Vistnes. Theopinions expressed in this paper are solely those of the author.

3 “The Commission has authority to review all mergers, acquisitions and takeover bids and other deals that can be defined as a‘concentration’, involving companies with a combined turnover worldwide in excess of C5,0 billion and European sales of at leastC250 million.” Commission press release IP/01/939. See also Article 1 (2) (a) and (b) of the European Merger Control Regulation,Council Regulation 4064/89 EEC of 21 December 1989 on the Control of the Concentration between Undertakings, 1990 O.J. (L 257)13, as amended by Council Regulation 1310/97 EC of June 30, 1997, 1998 O.J. (L 180) 1.

4 While the parties have appealed the decision to the Court of First Instance of the European Communities, a judgment is not expecteduntil sometime in 2003, too late to resurrect the prohibited transaction


23

THE PLAYERS

GE’s 2001 revenues exceeded $125 billion, and its businesses included everything fromplastics and television (NBC) to financial services, power systems, medical imaging, andlighting.5 In the arena of aviation, GE produces aircraft engines on its own (GEAE) andthrough CFMI, a 50/50 joint venture with SNECMA (a French company). This jointventure accounted for a large majority of GE’s engine sales, as CFMI is the exclusiveprovider of engines for Boeing’s most popular plane, the 737. CFMI engines also powerthe Airbus A320 family and the A340-200/300. GE’s own engines power the Boeing 777,767, and 747 planes. They also power the Airbus A300, A310, A330, and the not-yetlaunched A380 super-jumbo aircraft.6 In almost all of these cases, CFMI and GEAE competewith Pratt & Whitney (a division of UTC), Rolls Royce, or IAE (a PW/RR joint venture).

Honeywell started out in heating and environmental controls and over time developed a leadership position in aerospace.7 Honeywell’s position in avionics was enhancedthrough a series of mergers, most notably the purchase of Sperry Aerospace in 1986 and a merger with Allied Signal in 1999. Allied Signal was itself a leader in aerospace,the result of Allied Corporation’s purchase of Bendix in 1983 and merger with the Signal Companies in 1985. Along with avionics, Honeywell’s nonavionics aerospaceproducts include auxiliary power units (which gives power to the plane when on theground), starter motors, environmental control systems, aircraft lighting systems, engine accessories and controls, wheels and braking equipment. In 2001, nearly half of Honeywell’s $23 billion of revenue came from their aerospace division.

3 The case against the merger

To block a merger, the Merger Control Regulation requires the Commission todemonstrate that the proposed merger would lead to market dominance. According to European case law, dominance is defined as:

a position of economic strength enjoyed by an undertaking which enables it toprevent effective competition being maintained on the relevant market by giving it the power to behave to an appreciable extent independently of its competitors,customers and ultimately of consumers8

In this case, the claimed route to market dominance was unusual. It was not through the merger of competitors. Nor was it through a vertical integration of customer andsupplier.9 Given the depth and breadth of GE and Honeywell product offerings, there was

5 GE’s history goes back to Thomas Edison. General Electric was formed as a result of a merger between Edison General ElectricCompany and Thomson-Houston Electric Company in 1892.

6 The GP7000 engine, designed to power the Airbus A380, is a joint venture between GE and Pratt & Whitney.

7 Honeywell instruments helped guide Apollo astronauts Neil Armstrong and Buzz Aldrin’s landing on the moon.

8 In its investigation of the merger, the US Department of Justice found two significant horizontal overlaps, in US military helicopterengines and some heavy maintenance services for Honeywell aircraft engines and auxiliary power units. Both of these overlaps werefixed with a proposed divestiture. This proposed fix led to the US Department of Justice clearing the merger on 2 May 2001 (seehttp://www.usdoj.gov/atr/ public/press_releases/2001/8140.htm).

9 The very few areas of competitive overlap could have been resolved by divestitures. GE’s role in speculative aircraft purchases forleasing and Honeywell’s starter motor business did bring up some issues of vertical integration. We discuss the speculative aircraftpurchasing aspect of the merger is the case study on GECAS that follows.


24

a remarkable lack of overlap between the two companies. Instead, the focus of the merger review was on “portfolio effects” or horizontal integration issues of bringingcomplements together.10

The Commission’s case against the merger emphasised three linked points.

First, the Commission claimed that GE had a dominant position in aircraft engines forlarge commercial aircraft and Honeywell had a “leading” position in avionics and non-avionics areas for these aircraft.

Second, they claimed that the proposed merger would allow the new firm to bundlethese complementary products. The bundling strategy would lead to price discountsthat would give the firm an unbeatable advantage over its rivals.

The third leg of the argument was that this advantage would lead to the exit of rivalsand thus ultimately to strengthening the dominance of GE.

As explained in the EU Final Decision

Because of their lack of ability to match the bundle offer, these component supplierswill lose market shares to the benefit of the merged entity and experience animmediate damaging profit shrinkage. As a result, the merger is likely to lead tomarket foreclosure on the existing aircraft platforms and subsequently to theelimination of competition in these areas.11

The Commission was also concerned that GE’s aircraft leasing arm, GE Capital AviationServices (GECAS), might lever its purchases in a way to foreclose Honeywell’scompetitors.12 We discuss these issues extensively in the next case study.

To an economist, the Commission’s case was unorthodox. The concern was that underthe merger prices would fall - not that prices would rise. The idea that bundle discountscould be an anticompetitive strategy was novel to this case.13 A merger that created costsavings that resulted in lower prices would be permitted. But if the lower prices camefrom pricing efficiencies, this was viewed as anticompetitive. There was no discussion ofwhether the lower prices on the road to market dominance would or would not result ina net increase in the present discounted value of consumer surplus (i.e. increase consumerwelfare). Such considerations do not appear to be within the Commission’s mandate.

Even to a lawyer, the Commission’s case was unorthodox. The Commission did notdemonstrate that the combination would lead to dominance, as required by its mandate.Rather, it emphasised the theoretical potential for future anti-competitive behaviour.14

10 The Commission referred to this as a conglomerate effect. We prefer the term horizontal integration of complementors as itemphasises the specific nature of the relationship - namely that their two products are used together by a common customer.Honeywell’s customers are primarily the airframe manufacturers and airlines, not the engine maker. Thus Honeywell is primarily acomplementor to GE as opposed to a supplier. (See Brandenburger and Nalebuff (1996) for a more formal definition ofcomplementors.) Starter motors (and a few related components) are the exception to this rule.

11 Commission decision of 03/07/2002 declaring a concentration to be incompatible with the common market and the EEA agreement,Case no COMP/M.220 - General Electric/Honeywell (‘EC final decision’), at ¶355

12 A good summary of the Commission’s case regarding GECAS is presented in Giotakos et. al. (2001). See Pflanz and Caffarra (2002) foran excellent analysis of the unorthodox arguments made regarding the role of GE’s leasing company, GECAS.

13 Range effects had been considered in a number of prior EU cases, including Coca-Cola/Amalgamated Beverages and Guinness/GrandMetropolitan. These cases focused on distribution efficiencies rather than pricing strategies.


25

In this case study, we focus on the Commission’s arguments concerning bundling. (A separate case study looks at the GECAS aspect of the decision.) Not only is thisaspect of the case novel, it also has wide-ranging implications for other mergers. We present the Commission’s case for why bundling might create an antitrust issue, and then discuss whether the theory and evidence justified their conclusion.

4 Why bundle

We begin with a very brief review of bundling. Just because two items are sold as a package does not in itself mean that the items are bundled. The answer depends on how else and at what prices the items can be purchased. For example, if two itemsare only sold together and are not available separately, this is a case of pure bundling.The most general form of bundling is mixed bundling and this was the type of bundlingemphasised by the Commission. In mixed bundling, two items, say jet engines andavionics, are available both separately and as a package. What makes this bundling is thatthe package is sold at a discount relative to the individual items. If the package is simplypriced at the sum of its component prices, and these components are each available on an a la carte basis, then we do not call this bundling, as there is no strategic impact of the pricing.

The potential advantages of bundling were appreciated as early as 1838 by AugustinCournot. He considered the case where a customer needs to buy two goods that areonly of value when used together. Cournot used as an example the case of copper andzinc that are combined to make brass. Here one can think of the two goods as a jetengines and avionics. Bundling will be advantageous if each of these goods is sold by a monopolist.

Cournot’s insight is that two monopolists, acting independently, will set an inefficientlyhigh price. Were they to merge or coordinate their pricing, they would lower their priceand earn more money. The simple intuition is that the lower price of jet enginesstimulates sales of planes and thus avionics and this effect is not considered when thetwo goods are sold independently.

It is not surprising that the merging firms make more money. What is unusual here is that prices fall so that consumers are also better off. The merger leads to the happycircumstance in which all parties are better off, what is called a Pareto improvement.Thus the antitrust authorities should encourage such mergers.15 We return to this issuewhen we discuss policy implications.

14 Schmitz (2002) criticises the decision on this basis: “Although it is probably true that the new company would indeed have thepotential to bundle and it cannot be ruled out that at one point in time it might engage in this behaviour, using this potential toconclude that the merger would strengthen a pre-existing dominant position within the meaning of Article 2 European Merger ControlRegulation is questionable ... Describing the question of whether it is permissible to block a merger because of possible futurebundling as theoretical, hardly fits the impact it has. ...[T]he tool for this investigation is and must be Article 82 EC, not the EuropeanMerger Control Regulation.”

15 One of the ironies of this case is that if one took the view that GE/Honeywell each had a monopoly position then bundling wouldunambiguously improve welfare. The only possible source of harm would be on competitors. But to the extent that the firm does notface competitors, there is no harm done.


26

The Cournot example is the horizontal equivalent of “double marginalisation.” Each firmcauses a negative externality on the complementary products by raising its price. Whenthe two firms combine, they internalise this effect and lower prices. The simplicity ofCournot’s argument has led to general confusion about when the theory is applicable.

In some ways this result is very general. It does not depend on the specific form of the demand function nor the cost function. It does not require that the goods be perfectcomplements. But there are several hidden assumptions on which the result relies. Two, in particular, are of concern to us.

First, the basic Cournot model does not consider the impact of the merger on any otherfirms in the market. This is by design. In Cournot, the producers of copper and zinc werealone in the market. To apply this approach to the facts as interpreted by the Commissionin the GE-Honeywell merger, we need to consider how the results change when the twomerging firms are not alone in the market. Naturally, the results are more complicated.

There are now two reasons to cut price: market expansion and competition with rivals.Cournot looked at the price reduction as a way to increase the total market. Though thetotal demand for airplane engines can be expanded with a price reduction, the engines(along with avionics) are only a minority of the total airplane cost. The greater impact of a price cut is through the potential to gain market share from rivals.

Because there are rival firms, there will also be a response to a price cut. This responsemay offset the potential gain to the merging firms. Thus we will want to consider theequilibrium impact on the non-merging firms and on consumers to determine the overallsocial welfare implications.

A second reason why the Cournot framework may not apply to the GE-Honeywellmerger is that the basic result depends on an unstated assumption: that firms set asingle price in the market to all customers. This is a quite reasonable assumption for atypical consumer good, such as Microsoft Office. But it is not a reasonable assumptionfor the sale of large commercial products in which the two parties engage in extensivenegotiation as part of the sale process. If firms can price discriminate or negotiate witheach customer, then the advantage to bundling disappears.

Our focus here will be on the applicability of the Cournot model. There are othersreasons to offer a bundle besides curing double marginalisation. For example, even amonopolist seller of both A and B might be able to use bundled pricing as a way toimprove its ability to engage in price discrimination.16 This incentive to bundle is moreimportant if there is demand for the goods individually (as is the true case with copperand zinc) than when all customers buy both goods (as is typically the case with jetengines and avionics).17 We do not emphasise this aspect of bundling, as it did not play a role in the GE-Honeywell case.

16 See, for example, Adams and Yellen (1976); McAfee, McMillan, and Whinston (1989); Bakos and Brynjolfsson (1999).

17 There is still an advantage to bundling in a world where all customers buy an A-B package and the rival sellers of A (and B) areimperfect substitutes. This advantage only exists so long as the firm cannot perfectly price discriminate.


27

COMPETING AGAINST BUNDLES

We first extend the original Cournot model to cover the case where the two sellers face competition in the market. The results from this approach were at the heart of theCommission’s theoretical argument against the merger.

This approach is based on Nalebuff (2000). We consider the case with four firms in themarket. For simplicity, we can imagine that there are two jet engine manufacturers, GEand Rolls Royce, and two avionics manufacturers, Honeywell and Hamilton Sundstrand.

ASSUMPTIONS OF THE MODEL

In an attempt to demonstrate the effect of competition and not to create a realisticmodel of competition between the sets of firms, we assume that each customer hassome preference for one product over the other and that preference is uniformlydistributed over the interval [-1,1].18 Thus a customer with preference (0.2, -0.6) would bewilling to pay an extra 0.2 to buy a GE engine but would require a discount of 0.6 topurchase Honeywell avionics instead of Hamilton Sundstrand.

The pricing externality suggests that if GE and Honeywell can get together andcoordinate their pricing, they will have an advantage over Rolls and Hamilton, each actingindependently. But this gain may be offset by an increase in competition. If GE/Honeywellsell their product as a bundle and take share away from Rolls and Hamilton, theirexpected competitive response will reduce GE/Honeywell’s profits. Which effect dominates?To create a benchmark, we set the profits of each of the four firms to 0.50 in thecompetition prior to the merger.

Post-merger, GE and Honeywell bundle and profits are

∏GE/Honeywell = 0.91, ∏Rolls = 0.32, ∏Hamilton = 0.32

The increased competition effect dominates so that bundling reduces GE/Honeywell’scombined profits from 1 to 0.91. GE/Honeywell does roughly 50% better than the sumof the uncoordinated B firms. The market share moves from a 50:50 split to 63:37. But even with this gain in share, GE/Honeywell does about 10% worse than in the casewhere each product is sold in an uncoordinated fashion.

The explanation is that the bundle takes away enough market share from Rolls andHamilton so that the resulting equilibrium prices are low enough to make GE/Honeywellworse off.19 Thus, even though it leads to an advantage, there is no incentive to bundle.

Again, these numbers are for illustration and are not designed to fit the customerpreferences and product differentiation that exists in the jet engine and avionics industries.Furthermore, these results, adapted from Nalebuff (2000), consider the case of purebundling — the GE engines and Honeywell avionics were only available as a bundle.

18 The assumption of a uniform distribution over [-1, 1] is made for simplicity. It implies that all preferences between +1 and -1 areequally likely.

19 As the bundle grows in size, the gap in prices continues to grow and, consequently, so does the bundler’s market share. Once thebundle has four or more items, equilibrium profits rise for the bundling firm. See Nalebuff (2000).


28

To apply this model to GE-Honeywell, the results need to be extended to cover the caseof mixed bundling.

Rolls Royce presented to the Commission an extension of this model to include the caseof mixed bundling.20 While it is difficult to find a mathematical formula for the solution,the approximate equilibrium prices and profits can be found through simulation. The resultsare not identical to the pure bundling case, though they have the same flavour.

With mixed bundling, the GE/Honeywell bundle would be sold for a 19% discount belowthe pre-merger price, while the component prices for GE and Honeywell would each riseby 21%. Rolls and Hamilton would respond to this increased competition by loweringtheir component prices by 11%.

∏GE/Honeywell = 0.97, ∏Rolls = 0.40, ∏Hamilton = 0.40

Even with mixed bundling, GE/Honeywell would still sacrifice profits by bundling. Their combined profits fall from 1.00 to 0.97 or 3%. Although its profits fall, GE/Honeywellwould gain market share and lose less than their rivals. GE/Honeywell’s market sharewould rise from 50% to 55.4% while Rolls and Hamilton would see their profits fall by 21%.

MODEL LIMITATIONS

This model is obviously quite stylised. These parameters are in no way reflective ofexisting jet engine and avionics markets. The results about whether or not the bundlingis profitable can be reversed with relatively minor changes in parameter values ormodelling assumptions. For example, bundling becomes more profitable the more itemsare added to the bundle (at least for the case of pure bundling).21

Using a highly stylised model, the Commission reached the conclusion that economicincentives would lead a firm to engage in mixed bundling. The result that a firm thatbundles obtains an advantage over its rivals is a relatively robust conclusion. But whetheror not a multi-product firm has an economic incentive to bundle is a much more delicatefinding. And by the same token, so is the expected loss to the competition. Among otherfactors, it depends on the number of items in the bundle and the elasticity of totalmarket demand. The results presented rely on a specific distribution of preferences,namely a uniform distribution. It is not clear if the results are robust across differentdistributions of customer preferences.

The results also depend on the two goods in the bundle being of symmetric importanceto the consumer. Clearly the engine is more important than the avionics or even all of thepotential Honeywell components combined.

An aircraft engine might have a $15 million price tag while a piece of avionics could sellin the $100,000 range. Overall the avionics components add up to less than 5% of thetotal aircraft cost. The basic model of bundling has the two products being symmetric in

20 The non-confidential version is presented in Choi (2002).

21 While the biggest impact from lowering price comes from gaining market share, there is also the potential to expand the total market.As recognised by Choi (2002), even if this effect is small, it can be enough to make bundling profitable.


29

valuation. Nalebuff and Lu (2001) extend the earlier model to allow for asymmetry inimportance. In the examples considered, there appears to be little incentive to bundleand minimal impact on competitors. In fact, with pure bundling and enough asymmetry,bundling actually increases the profits of all the players in the market.

The Rolls Royce model only examines the two-goods, two-vendor case. Realistically, the number of avionics and nonavionics goods purchased is in the several dozens, andthe number of important vendors is at least a dozen. We need to understand not justwhether mixed bundling is attractive or not, but also how much more attractive it is asthe bundle size increases and what impact that increased bundle has on the market.

Thus before we try to use these type of models to make predictions about the likelyimpact of a merger, the models need to be robust enough to capture some importantelements of the real-world market.22 If one takes the results of this model at face value,it is not clear why there should be any antitrust concern. Average prices fall in themarket. Consumers are better off at the expense of firms.

One can make the argument that social welfare falls, but as Patrick Rey observed in hispresentation to the Merger Task Force this is truly an artefact of the model. As totaldemand is inelastic, the only change in social welfare is due to a change in which customersbuy which product. Since all customers start out with their ideal product (absent priceconsiderations) in the initial symmetric equilibrium, any change would lead to a fall insocial welfare. But this argument depends critically on the starting points being completelysymmetric. If, for example, the merging firms have a superior cost or product positionand, thus, a larger share, then the bundle discount can lead to increased efficiency.

The Commission’s problem with bundling was not with the immediate loss to socialwelfare but rather with the long-run (and even short-run) impact on competition. Similarto the argument against predation, the Commission believed that rivals would exit andthat GE-Honeywell would obtain and exploit a dominant position.

DYNAMICS

If one wants to consider the expected market dynamics, then there are other factorsthat should be taken into account. For example, is it realistic to imagine that theindividual competitors would not respond in any way? One option for those firms is to invest in product improvements or cost reductions. Another option would be to offer a competing bundle.

In our basic model, were Rolls and Hamilton to offer a competing bundle that would leadto a further reduction in profits, but it would also level the playing field. Firms may preferto be in a symmetric position relative to a rival. They may prefer a lower but level playingfield if they are worried that a bundler would use its profit advantage to position itselfbetter in R&D or gain other advantages in a repeated game.

22 But even assuming for a moment that all of these issues did not exist, work by Patrick Rey shows that even on its own terms theRolls Royce model was fatally flawed. In order to estimate the impact of the proposed merger, one needs to have coefficients on themodel. The problem is that there is only one observation to use to estimate these parameters. Rolls Royce used the existing marketshares. At best that would leave the model exactly identified. However, in this case the model was still underidentified. Thus theimpact of the merger would depend on the choice of some parameters for which there were no data to make an estimate.


30

Even if Rolls and Hamilton did not want to offer a competing bundle, their customerscould drive them to this outcome. Customers would stand to gain a great deal if theycould create a bundle-against-bundle competition. Customers are anything but passive in this market and would use their power to influence the nature of competition.

The Commission only considered the advantages of bundling when the competition wasselling their products individually. Whatever advantages may exist, they quickly disappearif the rival firms coordinate and offer a competing bundle. Given the level playing field forrivals and the advantages to customers, a firm that introduces a product bundle cannotexpect that rival firms will not offer competing bundles.

5 Negotiating bundles

Up to this point, we have been playing the game as defined by the Commission and theRolls Royce model as if it was relevant to the merger case at hand. We have shown thateven on its own terms, there are sufficient flaws in the application that we would not beable to rely on its conclusions.

Here, we suggest that there is a much bigger flaw with the entire modelling approachsuggested by Rolls Royce. All of the results on bundling, starting with Cournot, dependcritically on the assumption that there is one price to customers in the market. This is avalid assumption for almost all consumer products. However, this assumption does notapply to the aerospace industry.

Customers don’t pay list prices for jet engines or avionics. Airplane customers are oftenlarge and powerful. A vendor cannot ignore a major airline that asks for a better price.Nor are vendors uninformed as to their customers’ preferences. The vendors in thismarket spend a great deal of resources getting to know their customers. Vendors takeinto account previous purchases as well as technical performance differences betweentheir products and those of competitors. Going into a competition a vendor has a goodidea of where it stands and by the end of the competition, it has a very good idea.

The result of this buyer power and vendor information is that prices are negotiated, notset by the seller. Examples of transaction prices confirm the fact that different buyerspay different prices. Every deal is negotiated, and the price is customised to thespecifics of the situation.

In a world where firms negotiate prices with customers and do so with perfectinformation, the combination of two complementor firms, such as GE and Honeywell, iscompletely neutral. To see why, take the case where a customer has a preference for theGE engine and the Honeywell avionics. For illustration, imagine that the advantages areequal to (0.2, 0.3). In this case, GE and Honeywell would be expected to win bothcompetitions, whether the items are sold separately or if the firms merge and theproducts are sold in a bundle. Let us compare these two cases.


31

Before conceding defeat, both Rolls and Hamilton would be willing to price down tomarginal cost. In the case where the engines and avionics are sold separately, GE wouldbe able to negotiate a profit up to 0.2 on its engines - its profit equals its advantage inthe market. Honeywell could negotiate a profit of up to 0.3 on the avionics.

If the goods were sold as a bundle, then GE/Honeywell would be able to negotiate aprofit of up to 0.5 on the bundle. Thus the sales and combined profits are identical in thetwo cases.

Nothing is different when Rolls and Hamilton has an advantage in one (or both) of thegoods. Take the case where the customer has a preference for GE/Honeywell productsof (-0.2, 0.3), so that the customer actually prefers the Rolls engine to the GE engine by0.2. With individual pricing, GE would lose out to Rolls while Honeywell would still winover Hamilton (and make a profit of 0.3).23 If the merged GE/Honeywell firm tried to sellthe two products as a bundle, it could do so, but only at a profit of 0.1. This would beworse than selling just the avionics at a profit of 0.3.

These two examples are quite general. The point they illustrate is the following: When the customer type is known and prices are negotiated, bundling can never lead tohigher profits. If the customers would have made all of their purchases from a single firm,bundling has no impact - on customers, prices, profits, or efficiency. If customers wouldprefer some products from Rolls and others from Honeywell (or vice versa), then thecombined firm will continue to offer the individual goods at their pre-merged prices.Forcing a bundle on the consumer can only lower firm GE/Honeywell’s profits. In effect, itwould have to subsidise its disadvantage using profits it could have earned from productswhere it is strongest. This is no different from selling individual components at a loss - astrategy it can do but would choose not to, even without bundling.

The perfect information negotiation model is designed to capture the basic nature ofcompetition in this market. But, like all models, it presents a simplified description of themarket.24 While vendors are well informed about the customer, their information is notalways accurate. Firms can still negotiate prices even with good but imperfectinformation as to customer preferences.

The conclusion from this negotiation model is not a narrow result. The mathematics ofthe more realistic cases becomes more difficult, but simulation results suggest thatbundling has little effect when vendors have good but not perfect information.

As one example, imagine that the firms don’t know the customer’s exact preferencesbut which firms the customer prefers.25 Thus instead of saying the customer is (0.2, 0.3)we would say the customer is (+, +), indicating that the person has some positive butunknown level of preference for each of the GE engine and Honeywell avionics.

23 This example assumes that preferences for engines and avionics are of similar importance. If differences in avionics are lessimportant, then a typical consumer might have preferences (0.2, 0.03) - in this case, engines are 15 times more important thanavionics. Hence, the result extends directly to cover cases where the different products are not of equal importance.

24 An important simplification is that we are considering only a static snapshot of competition. We do not take into account issues ofbrand loyalty, switching costs, dynamic R&D incentives, or predatory strategies that require recoupment.

25 For more specifics on this approach, see Nalebuff and Lu (2001).


32

Consider a customer who is known to prefer both GE and Honeywell products, a (+,+)type. In this case, bundling offers a small advantage. A firm that knows it has anadvantage in all components can use a bundle to do a better job of price discrimination.Market efficiency increases. Competitors’ profits fall, but their profits in this case werevery low to start with.

In contrast, bundling is not profitable when the merging firm is at a disadvantage in allproducts (-,-). Here we also note that this is the case where rivals (Rolls and Hamilton)have the highest profits. Thus bundling has the lowest impact when rivals make the mostmoney.

Pure bundling or tying would be counter-productive in the two cases where the mergingfirm (GE/Honeywell) is better in one component and worse in the other, (+,-) or (-,+). If the merging firm employs mixed bundling, then the majority of consumers do not takeadvantage of the bundle, and there is again a small impact of the mixed bundle.

Averaging across these four cases, the net impact of the mixed bundling strategy isreduced by 50%. This is in line with the improved information. Half of the uncertaintyhas been removed in the sense that each firm knows if it is ahead or behind, althoughnot by how much.26 Nalebuff and Lu (2001) show that with even better information, the impact of mixed bundling is even smaller.

The distribution used to generate these simulation results provides results for four typesof customers ((+,+), (+,-), (-,+), and (-,-)), but that leaves wide open the question of whatis the proper level of information about the customer and what is the distribution of theremaining degree of uncertainty. The tightness of the distribution is a proxy for thequality of information in any particular negotiation. The wide margin variations that weobserve are indicative of high-quality information, but this is not something that has beenempirically measured and calibrated to the model.

In summary, it would be possible to employ this negotiation approach to model thepotential impact of bundling as a result of the GE/Honeywell merger. The model wouldrecognise asymmetries in the products being bundled as well as the negotiationenvironment with imperfect information. Doing so would require an investigation as tothe quality of information and the distribution of consumer preferences. Neither of thesetasks was ever considered. Since the size of the effect is directly related to the degreeof imperfect information, it is incumbent on the Commission to incorporate negotiationand uncertainty into their model. They have not done so. Thus we conclude that theCommission does not have a sound theoretical basis on which to conclude that bundlingcan lead to market dominance.

26 Another way of putting this is that each firm knows which half of the line the customer is in and thus the range of uncertainty hasbeen reduced by half.


33

6 Empirical evidence of dominance and bundling

Ultimately, the Commission’s case against the proposed merger rested on a claim ofmarket dominance. The empirical starting point in this argument was that GE already hada dominant position in aircraft engines for large commercial aircraft. Recall that marketdominance means that a firm can act independently of its rivals and its customers.

It almost follows from the definition that a firm without a commanding market sharecannot have a dominant position in the market. The Commission presented itscalculation that GE had a dominant position with a 52.5% market share of the installedbase of engines on large commercial aircraft still in production:27

MARKET SHARES:

AIRCRAFT ENGINES INSTALLED BASE

GE PW/IAE RR/IAE

Narrow Body 51 22 27

Wide Body 54 31 15

Overall 52.5 26.5 21

Planes still in production leaves out planes still in service but no longer produced. While new engines are obviously no longer sold to planes no longer being built, thisperspective misses the spare parts market. With spare parts an engine can be “sold” up to ten times over its working life. Pratt & Whitney, in particular, has a large annuitycoming to it from selling spare parts to planes in service that are no longer beingproduced. Using the planes-in-service approach, GE’s market share falls to 41%.

The calculation of market shares for PW (Pratt & Whitney) and RR (Rolls Royce) includethe engines of IAE, a joint venture between Pratt & Whitney and Rolls Royce. IAE’smarket share is split evenly between PW and RR. In contrast, all of the market share ofCFMI, the 50/50 joint venture between GE and SNECMA is attributed to GE.28 If CFMI iseliminated from consideration and we consider market share on planes in service, thenGE’s market share falls to 10%. If we attribute half of CFMI to GE and half to SNECMA,then GE’s market share is still only 28%. Even if we use the planes in production, GE’smarket share counting only half of CFMI to GE is 36%.

27 The data is of 31/12/2000. The table is reproduced from European Commission decision in Case No. COMP/M.2220 - GeneralElectric/Honeywell, paragraph 70.

28 The Directorate-General Competition justified this calculation: “Although in legal terms GE and SNECMA jointly control CFMI, the onlymeaningful attribution of market shares for the purposes of analysing the transaction could only be made to GE, to the extent thatSNECMA is not an independent supplier of civil jet engines for large commercial aircraft. The analyses of the joint venture and ofSNECMA’s participation in other GE engine programmes indicated that SNECMA would act jointly with GE as a profit-maximisingentity.” Giotakos et. al. (2001).


34

In any of these calculations other than that of the Commission, GE’s limited market sharewould practically preclude it from having a dominant market position. Which definition is correct?

There is no one correct definition of market share made in the abstract. The issue shouldbe market share measured for what purpose. For example, if the purpose is to evaluatethe firm’s financial resources to continue investing in this business, then all of itsrevenue streams are relevant. Thus PW can use earnings from its entire installed base(not just planes in production) to finance new investment. GE only gets half of therevenue from CFMI. Thus for this purpose it would seem that the 28% number is mostappropriate as it reflects GE’s revenue share of today’s market.

To understand future financial viability, one might also want to examine pre-orders on next-generation planes, those that have been launched but are not yet in production(such as the A380, B777x). It is common for airplanes to be ordered in advance ofproduction. As of 2001, Rolls Royce had a 40% share of the engines on these next-generation planes, while GE and CFMI’s combined share was second with 38% and PW was third with a 21% share.

If market shares are used to provide evidence of market power, it would seem that most of CFMI’s share should not be included in GE’s share. This is because the greatmajority of CFMI’s sales are not in a position to exercise any market power. Recall thatCFMI is the exclusive provider of engine to the Boeing 737 plane.29 Prior to entering this exclusive relationship, Boeing realised that its customers would be in an untenablebargaining position if CFMI were the sole provider. Thus Boeing prenegotiated a dealwith CFMI as part of the exclusivity contract.

For most engines, the airline purchases the engines separately from the plane. Having decided upon a 747, the airline can then put the engine order out to bid betweenGE, PW, and RR. But in the case of a 737 purchase, the customer negotiates withBoeing and the engine is included in the price. CFMI does not have the ability to controlengine pricing on these orders. For that reason, a more appropriate measure of GE’smarket share used to measure market power would be its share of engines excluding itsexclusive-contract sales.30 This corrected share is about 10% or 20% depending onwhether one looks at planes in service or just those in production.

If market shares are used to understand the potential for bundling then, again, it wouldseem that CFMI’s sales should not be attributed to GE. This follows for two reasons:

First, for the engines offered to Boeing at a predetermined deal it is almost impossible to employ a bundled pricing strategy. If the customer has already bought the Boeingplane and CFMI engine, there is no gain in providing a retroactive engine discount for the purchase of Honeywell components. The company would do just as well to offer the discount directly on the Honeywell parts.

29 In addition to CFMI’s exclusive sales contract for the Boeing 737, its CFM56 engines also power the Airbus A320 family and A340 longhaul.

30 There are some GE exclusive sales, such as for the two new “longer range” (LR) versions of the Boeing 777. Here, again, the engineis sold at prenegotiated terms. One would want to take out all of the exclusive engine sales in calculating a firm’s ability to actstrategically.


35

The only way even to offer a bundle discount would be to promise a future discount on Honeywell components if the 737 plane is purchased. This would only make sense ifthe discount would increase the 737 plane sales. (If the sales of 737s are unchanged,then there is no increase in demand for the CFMI engine, and hence the discount mightas well be applied directly to the Honeywell components.)

There are practical problems with this approach. The cost of Honeywell components issmall relative to the plane and engine. Thus there is little ability for the tail to wag thedog. Second, the sale of avionics and other Honeywell components is done at a futuretime at a negotiated price. One party can’t promise to give the other party a “betterdeal” in a future negotiation, as there is no baseline against which to measure whatmakes a better deal.31

Second, bundling would be further complicated by the fact that CFMI is a joint venture.Thus the decision on offering a bundle price would have to be made by both GE and itspartner SNECMA. By its charter, CFMI is always led by a SNECMA representative. The joint venture nature of the relationship complicates CFMI’s ability to offer a bundle,as SNECMA has no incentive to help sell Honeywell avionics.

GE’s ability to act strategically so as to gain market share is practically restricted to itsown engine sales, and this is roughly 20% of the market for planes in production. A 20%market share does not put a firm in a position where it can act independently of its rivalsand customers.

However one measures shares, there is no disagreement regarding the extensive use of bidding competitions, the emphasis on market intelligence, and the very dynamicnature of shifting market shares. But the same facts were viewed very differently on thetwo sides of the Atlantic. In the US, the fact that GE had won several of the recentengine competitions was viewed as evidence of competition at work. In Europe, theserecent wins were viewed as evidence of dominance.32

A contributing factor to GE’s alleged dominance in engines was the company’s uniquefinancial strength. Excerpts from Giotakos (2001) demonstrate this perspective:

GE Capital offers GE business enormous financial means almost instantaneously andenables GE to take more risk in product development than any of its competitors ...GE has also taken advantage of the importance of financial strength in the industrythough the use of heavy discounts of the initial sale of the engine. ... [T]hanks to itsfinancial strength and incumbency advantages as an engine supplier, GE can afford toprovide significant support to airframe manufacturers under the form of platform-programme development assistance that competitors have not been historically in aposition to replicate. ... Unlike any other engine manufacturer, GE can afford to pay forexclusivity and capture aftermarket, leasing and financial revenues.

31 To help make this clear, consider the following hypothetical bundling offer: if “New Air” purchases a 737 plane, GE will offer a $1million rebate good towards the purchase of Honeywell avionics. When it comes time to negotiate the price of those avionics,Honeywell and the customer both factor in the $1 million discount when setting the price. Thus if the negotiated price would havebeen $4 million, the new negotiated price would be $5 million, which achieves the $4 million post-rebate price.

32 For a more detailed comparison of the US and European approaches, see Patterson and Shapiro (2001) and DOJ (2001).


36

Patterson and Shapiro (2001) show the dangers in the approach. They look at as thesesame activities as procompetitive. Taking risks leads to innovation; discounting benefitscustomers. In the US, entrenchment of a dominant firm is a discredited theory and is nolonger grounds for challenging non-horizontal mergers (DOJ, 2001):

Challenging a merger because it will create a more efficient firm through economiesof scale and scope is at odds with the fundamental objectives of antitrust law. Andthere is no empirical support for the notion that size alone conveys any significantcompetitive advantage that is not efficiency-related.

Whether or not GE was starting from a dominant market position, the Commission wasconcerned that this merger would allow GE to extend its dominant position to Honeywellproducts. The proposed route to this dominant position was through bundling. To whatextent was this theoretical concern supported by evidence?

While the theory suggests that bundling is unlikely to be an important factor, thereremains the empirical question: do we see much evidence of bundling in aerospace? At first glance, the answer seems to be yes: many bids are multi-item bids. Companiessuch as Honeywell and its competitors often make a bid to supply a long list ofcomponents. This led several observers and even industry participants to conclude thatbundling was a feature of this industry.

However, the long list of items is also broken down into prices for each individualcomponent. And the component prices add up to the “bundle” or package price. If thereis no discount then we don’t consider this to be bundling.

Compare this situation to Microsoft Office. Microsoft’s July 2001 list price for Office XPProfessional was $547. One could also buy the components separately, but onewouldn’t. Word, Excel, PowerPoint, Access each cost $339 separately and Outlook wasa bargain at $109. The total adds up to $1,465. The software package comes at a 60%discount compared to the individual items.

In the cases cited by the Commission, the claimed bundle discounts were much smaller- by an order of magnitude. The claimed bundle discounts were also much smaller thanthose predicted by the Rolls Royce model. Even more importantly, these discounts wereevidence of negotiation and not of mixed bundling.

In cases where the Commission attempted to document examples of a mixed bundlediscount the analysis failed to distinguish between a discount and a discount conditionalon buying a package. As a result, several of the cases cited actually demonstrated theabsence of mixed bundling.

Even if a discount were offered for buying a package, this does not mean that the discountwould not be applied to individual items. In general, this would be a matter of speculation.However, in some of the cases cited by the Commission, the alleged mixed bundle failedto induce the customer to buy the entire package. Thus it is possible to observe whatprices were paid and whether or not the discount was in fact conditional on buying the


37

entire package. In the cited cases, the customer was not penalised for breaking apart thebundle. Whatever discounts the customer was offered for purchasing the bundle wereapplied on a proportionate basis to the partial bundle that was ultimately purchased.

The evidence of bundling provided by the Commission undercut its own argument. If bundling was such an anti-competitive tool, then why did it fail to get customers topurchase products for which the firm was otherwise at a disadvantage? The fact thatcustomers were able to get any discount offered without having to purchase the bundleconfirms the industry perspective that it is a mistake to offer a bundle discount as thiswill simply end up coming back as a discount on whatever the customer ends up buying.

If bundling is to be a matter of concern, we should see contracts that are won, not lost.And these contracts should offer a substantial discount for buying the entire packageover a la carte purchases.

7 The (im)practicality of bundling engines

The theory suggests that bundling will not lead to an advantage with negotiated prices.The evidence suggests that bundle discounts are not prevalent, if they even exist. But perhaps the combination of GE and Honeywell would create a new opportunity to offeran engine and avionics/nonavionics bundle. Even if this were desirable, there are institutionalfeatures of the airplane purchase process that make this type of bundling impractical.

We have already discussed the problems of bundling CFMI engines with Honeywell’savionics (and other nonavionic) equipment. These include the joint venture withSNECMA and the pre-determined engine price for 737s. An additional factor that makesit impractical to bundle engines with other components is the timing of purchases.Typically the engine choice is made well in advance of other components, such as avionics.

To see why this is a problem, consider how bundling would have to work. Once GE haswon the engine competition, there is no incentive to give retroactive discounts onengines if the customer would also buy Honeywell avionics. This would be no differentthan giving a discount on the Honeywell avionics directly.

In order for there to be any possibility of bundling working, it must be the case that at thetime the engine selection is being made the customer is led to believe that by choosingthe GE engine there will be a better price on Honeywell avionics. For example, the customermight be given a 10% off discount for their later purchase of avionics if they have a GEengine. While this story is possible in some industries, it is not applicable to avionics.The reason is that all prices are negotiated so that a discount off the list price has no bite.One person can’t promise to give the other party a “better deal” in a future negotiation,as there is no baseline against which to measure what makes a better deal. Here again,we see the importance of taking into account whether prices are fixed or negotiated.

The bundling of avionics components would seem to be much more practical thanbundling engines and avionics. Indeed this issue was considered only two years earlierwhen the commission approved the merger of Allied Signal and Honeywell. While there


38

was some dispute as to whether bundling exists at all, even the Commission did notargue the Allied Signal-Honeywell merger had led to widespread bundling. From theCommission’s perspective this was only because the merger was too recent and thusthe impact of bundling had not yet been felt.

8 Evidence that bundling would lead to exit by rivals

At the same time the Commission was arguing that the Honeywell-Allied Signal bundlingeffects were slow to arise, the Commission took the stance that the GE-Honeywellmerger would lead to nearly immediate exit or marginalisation of rivals.33

Is exit (or marginalisation) really likely in this business? Here the Commission seemed to rely on the dire warnings of some competitors. Of course, these competitors are not disinterested parties. They would hope to block the merger or purchase variousHoneywell “jewels” that the Commission would require to be spun off.

Nor was there evidence provided that the rival firms were in any danger of exit. To thecontrary, evidence was presented that showed the long-term viability of the rival firms.For example, the stock market’s reaction did not anticipate the financial vulnerability ofaerospace firms. From the time of the announced merger to the time of the EU hearing,almost all of the rival firms had a gain in stock price that exceeded the S&P index. The firm that under-performed the index was GE.

One would not expect to see aerospace firms concede defeat quickly. Airframes arelong-lived. The typical plane is on the market for 25 or more years. Thus a contract gainedtoday (or ten years ago) would provide a long stream of profits down the road. Hence, evenif firms were precluded from new contracts, they are not going to disappear any timesoon. In fact, the next plane for which the engine choice has not been made is theBoeing Sonic Cruiser, which is not set to debut until 2007 or later (if it is built, at all).

In the aerospace industry there are several large players who have an interest inmaintaining competition. Military purchases play this role. Both Rolls Royce and P&Whave multi-billion dollar annual revenues from military contracts, and there is a substantialamount of spillover between civilian and military work. While it could be argued that no single airline would be eager to provide the public good of maintaining competition,Airbus and Boeing (each of which has about a 50% market share and each of whichworks closely with airlines when making purchase decisions for the aircraft it designs) do have this incentive.

9 CONCLUSIONS

In the Statement of Objections, the Commission presented a theory of bundling that wasbased on the premise that the merged firm would be able to bundle and would have arational economic incentive to do so. The result of this bundling would be prices so lowthat competitors will be foreclosed.


39

When closer scrutiny was given to this argument, it turned out not to apply to themarket for aircraft engines, avionics, and non-avionics products. The Commissionsubsequently abandoned its original approach.

The various economic analyses have been subject to theoretical controversy, inparticular, as far as the economic model of mixed bundling, prepared by one of thethird parties, is concerned. However, the Commission does not consider the relianceon one or the other model necessary for the conclusion that packaged deals that the merged entity will be in a position to offer will foreclose competition from theengines and avionics/non-avionics markets.34

Instead, it based its decision on a new, dynamic, theory: Foreclosure of competitorswould occur as a result of predation accomplished through the cross-subsidisation ofbundled sales.35 As explained in the EU’s Competition Policy Newsletter (Giotakos, 2001)

...thanks to GE’s strong generation of cash flows resulting from the conglomerate’sleading positions on several markets, following the merger, Honeywell would havebeen in a position to benefit from GE’s financing surface and ability to cross-subsidiseits different business segments, including the ability to engage in predatory behaviour.

There may be a simple explanation as to why reliance on the Rolls Royce model was soattractive to the Commission. If, in fact, a merged firm would make more money byreducing prices when selling complementary products, then there is no need to estimatethe cost of predation (as there is none) and there is no need to estimate recoupment (asnone is required). Moreover, the merger becomes the proximate cause of the pricediscounting and so there is now a reason to think that this type of economically rationalpredation will occur post-merger even if it does not occur pre-merger.

Against this background, the Commission may have been disappointed to discover thatthe economic models upon which it had relied - the model advanced by Rolls Royce, andeven the basic Cournot complements model - were unsuited to the task.

Although the Commission abandoned the original model, it did not replace the flawedmodel with another model. Rather, the Commission switched to a dynamic theory ofpredation. But, it did not carry out the steps required to establish the facts necessary to support that theory.

In the end, the Merger Task Force did back away from its economic theory of bundling.But it did not back down from its conclusion that bundling was a reason to block themerger. The decision in this case presents a challenge for the role of economic analysisin the design and implementation of antitrust policy.

33 If the exit or marginalisation were to occur more gradually, then one would want to take into account the benefits gained by customersduring the period of lower prices. Such a calculation was never made.

34 Paragraph 352 of the Decision.

35 Predation was never seriously discussed in the Statement of Objections: there is only a single mention of predation in the SO,unrelated to the bundling theory, and this occurs in a footnote.


40

VGECAS and the GE-Honeywell merger: is the Archimedeantheory of leverage Kosher?1 Introduction

In this case study, we return to the GE-Honeywell merger to look at a second leg of the European Commission’s case against the merger. Here we examine the role of GECapital Aviation Services (GECAS) as a speculative purchaser of new airplanes for use in its aircraft leasing business.1

The issues raised by GECAS are relevant to understanding portfolio effects more broadly.What is the ability of a large firm, in this case a large customer, to throw its weightaround so as to distort competition in favour of its related businesses? We examine theGECAS case both in its details and its broad scope. The details illustrate the delicatenature of how to demonstrate market power and portfolio effects; the broad scopeallows us to examine how a theory of leverage works in practice.

The case begins with the fact that GECAS has (with one exception) only ordered newairplanes that come with GE or CFMI2 engines. What is in dispute is the interpretationand implication of this policy. Those challenging the merger took the view that GECAS’sfavouritism was costly and therefore an indication that GECAS was up to something. We will explore various motivations for favouring internal purchases so as to betterunderstand its legitimacy. The specific concern was that upon completion of the merger,the favouritism would be extended to Honeywell components, and that this would tipthe entire market towards Honeywell and thereby foreclose rivals.

What made this aspect of the case particularly controversial is the charge that GECASmight be capable of foreclosing rivals even though it had by all counts a market share of less than 10% and a share as low as 5% by some measurements. Thus a new theory

1 Barry Nalebuff assisted GE-Honeywell with analysis and responses to the Merger Task Force proceedings discussed in this case. The views expressed in this case study are solely those of the author and should not be attributed to GE, Honeywell, or any otherorganization. This case study follows and expands upon the arguments that are very well presented in Lexecon (2001), Pflanz andCaffarra (2002), Giotakos et. al. (2001), and Reynolds and Ordover (2002).

2 CFMI is a 50-50 GE/SNECMA joint venture.


41

of market power would be required. The resulting argument, an “Archimedean leveragetheory,” as developed by Robert Reynolds and Janusz Ordover (2002), was indeed novel,even ingenious. However, we do not find it persuasive.

We present the Reynolds and Ordover argument and then consider whether it isplausible on theoretical grounds. If the CEGAS purchase policy has any effect at all, isthis effect long-lasting or would it be offset by the equilibrium reactions of other players?Is it likely that post-merger GECAS would extend its favouritism to planes equipped with Honeywell aircraft components? We explore countervailing factors which suggestthe merger could lead to a market bias against Honeywell and that GECAS would find it practically impossible to limit its purchases to GE-powered planes equipped withHoneywell components wherever possible.

We then consider whether the market leverage theory is plausible on empirical grounds.Using the same data as Ordover and Reynolds, we find that the data is entirelyconsistent with the view that GECAS has had no share-shifting effect on the market.This interpretation follows directly if GECAS took most of its market share from airlinesthat would have otherwise ordered directly from airframers.

Given the very speculative nature of the theoretical argument applied to engines, the weakconnection to Honeywell components, and the inconclusive empirical support, it isremarkable that the GECAS controversy played a significant role in blocking the merger.

2 Aircraft leasing and a history of GECAS

The airline leasing business can be thought of as three activities. The first is the sale-leaseback agreement. Here, an airline that owns a plane will sell the plane to a financialservices company and immediately lease the plane back. This transaction changes thebalance sheets of the two companies and also may change who gets to take thedepreciation allowances.

This is a large part of GECAS’s business, but was not relevant to the merger case.GECAS engages in sale-leasebacks for all planes, not just those with GE/CFMI power.3

Second, since it is the airline and not the leasing company that orders the plane, theairline is the party that chooses all the components that go on the plane.

A second activity is the speculative purchase of used planes. Just as a car dealer maytake trade-ins or even purchase used cars for resale, airplane leasing companies helpmake a liquid market in second-hand airplanes. And just as a car dealer often has abroader selection in its used car options, GECAS has a large number of used non-GEpowered planes in its fleet. GECAS’s activity in used planes was not central to the mergercase, since GECAS has no say in determining the components that go on the plane.4

3 For ease of exposition, hereafter we refer to GE/CFMI power as GE power.

4 There was some discussion as to whether the GECAS purchase of used planes was done in a way to help an airline swap its fleetfrom non-GE power to GE-powered planes. This was a side issue at best.


42

The focus of attention was on the speculative purchase of new aircraft. In this business,a leasing company buys a new plane from an airframer before it has a customer towhom it can lease it. Thus, the purchase is speculative. The leasing company will makemoney if it can find a customer to whom the plane can be leased at an attractive price. It will lose money if it has to take delivery of a plane that cannot be put into service orcan only be leased with a heavy discount.

One reason why this business exists is because of the long lead time in ordering planes.If the airline customer could simply order a new plane from the Boeing showroom andtake it home that night, then the transactions could be of the sale-leaseback variety. But as there is often a several year backlog for a new plane delivery, an airline may go toa leasing company to obtain one of the planes already in the queue.

GE is a relatively recent entrant into the speculative purchase of new aircraft for use inits leasing business.5 Its first speculative orders were placed in 1996. Here, the GECASpurchase pattern was distinctive. It favoured the speculative purchase of planes with GE power. In its first five-years of purchases, it ordered planes with a total of 1192engines of which 16 engines were not GE or GE’s joint venture, CFMI.6

As can be seen in Figure 1 below, the years 1996 through 2000 were a time of solidgrowth for the entire leasing industry.7

FIGURE 1

TOTAL ENGINES (ALL SPECULATIVE ORDERS BY LEASING COMPANIES)

5 GE acquired the aircraft leasing business of Guinness Peat out of bankruptcy in 1993.

6 These 16 engines went on 8 Boeing 757 planes, for which GE engines were not an option.

7 The data source for this chart and all other figures is BACK Aviation, 2001. The orders include engines for large commercial aircraft andlarge regional aircraft.

0

200

400

600

800

1000

1200

1400

1600

91 92 93 94 95 96 97 98 99 2000

Total Engines


43

The growth during this five-year period came from two sources. First, this was a periodof solid earnings for the airlines. Second, there was a shift towards speculative leasing. The speculative leasing market share of planes (weighted by engines) rose from justabove 10% in the first half of the 90s to around 20% in the second half of the 90s.

3 Market shares

As in most antitrust cases, the first controversy was over how to measure market share. By all accounts, GECAS has a market share of no more than 10%. Normally, this wouldbe enough to stop the investigation of foreclosure of rivals. But in the novel theory reliedupon by the Commission, even a small share of purchases could matter.

According to the European Commission, GECAS is the market leader:

With around 10% of the total purchases of aircraft, GECAS is the largest purchaser of new aircraft, ahead of any individual airline. ... GECAS is also the market leader in terms of jet aircraft on orders and options with a total backlog of 796 jet aircraft at the end of 2000 (535 for ILFC).8

It would be quite unusual to count options as part of a market share. Counting planescan also be misleading. This misses the fact that a large regional jet costs only about 1⁄8ththe price of a 747.9 Both issues are relevant. Out of the 796 GECAS orders and options,150 were orders for regional jets and another 300 were options on regional jets.

To better understand GE’s size in this market, we look at market shares in terms of engines,not planes. There is not a one-to-one correspondence between engines and planes assome planes come with two engines, others with three, and others still with four.

In the period 1996-2000, GECAS’s firm and pending orders totalled planes with 1,192engines. This put GECAS into second place among leasing companies. The largest playerwas International Lease Finance Corporation or ILFC, with a total of 1,212 engines. The other leasing companies accounted for another 1,346 engines. The airlines accountedfor most of the orders. Their total was 12,544 engines over this period. Thus GECAS marketshare counted in terms of number of engines ordered was 1,192/(12,544+3,750) = 7.3%.10

Even this 7.3% share number substantially overstates GECAS influence as it counts allengines as equal. Engines differ widely in cost depending on the size/thrust of theengine so that an engine for a regional aircraft is only a fraction of the cost of an enginefor a large, wide-body aircraft such as an A380 or a Boeing 777. In dollar terms, GECAShas a lower market share as fully 1⁄3 of the GECAS total were regional jet engines.(GECAS was the only leasing player of any size in the speculative purchase of regionaljets.) Correcting for this we find that GECAS’s share in dollar terms is approximately 6%.11

8 Paragraph 122 of European Commission Decision in Case No. COMP/M.2220. ILFC stands for International Lease Finance Corporation.

9 The list price on a large regional jet is around $27 million compared to a little over $200 million for a Boeing 747.

10 Other leasing companies accounted for 15.7%, and airlines account for the remaining 77%. Note also that these numbers excludeengines for smaller regional jets (less than 70 seats). While leasing companies have not made speculative purchases of smallerregional jets, this market is relevant to the issue of whether Honeywell’s rivals could be foreclosed and thereby driven out of business.Including the smaller regional jets would have expanded the airlines’ share and reduced GECAS’s share of total new orders.

11 Note that if market share were measured in terms of dollar value of planes a similar result would be obtained as regional jets are lessexpensive than narrow-body and much less expensive than wide-body jets.


44

GE further argued that the CFMI engines should only count as 50%. Recall that CFMI isa joint venture between GE and SNECMA. As GE only earns half of the profits on anyCFMI sale, their incentive to generate an engine sale is half of what it would be on a 100%GE engine. Making this correction would bring the GECAS market share to below 5%.

As much of a stretch as it is for a firm with a 10% share to foreclose a rival, it becomesan even greater stretch for a firm with 5% or 6% market share.

Market shares are, of course, backward looking. Are these shares indicative of futuremarket shares? Here again, GE argued that the GECAS share was inflated relative to itslong-run position. The reason is that GECAS had, as of the end of 2000, placed two largeorders, one in 1996 and one in 2000. These orders were designed to build up inventoriesand hence not indicative of annual purchases. In 1996, GECAS ordered planes with 330engines. Having built up its inventories, the next year, in 1997, GECAS’s orders totalledonly 16 engines. In 1998 the total increased to 110 and to 118 in 1999. In 2000, thenumber jumped up to 618. This last number reflects the fact that in 2000, GECAS madelarge multi-year orders to enable it to undertake the speculative leasing of regional jets.Of the 618 engines ordered in 2000, 300 were building up inventories in this new regionaljet leasing business. If we take the two large orders and evenly spread them out over 4years, one backwards and two forwards, then we find GECAS market share is 5.5%,even counting regional jet engines equally with narrow-body and wide-body jet engines.12

We will return to discuss the market data when we look for any evidence that GECASwas able to shift the market in favour of GE/CFMI engines. Now, we turn to theArchimedean lever or tipping theory.

4 How does the leverage work?

The basis for the Reynolds and Ordover argument is that GECAS will be able to tip themarket in favour of GE/Honeywell products. We find their argument fits some businessdecisions very nicely. In particular, their Archimedean lever explains the prevalence of Koshercertification. It is not that the Reynolds and Ordover theory is never right, but rather thatit does not apply to the choice of aircraft components. Thus we first explain how theleverage theory applies to decisions such as the choice of Kosher certification and thenwe show how the certification decision is different from the choice of aircraft components.

Observant Jews will not eat food that has not been certified Kosher. While the populationor market share of observant Jews is small, under 1%, a typical manufacturer will still go through the expense of Kosher certification (when it is possible to do so). The reasonis simple. If the food is not certified, then the entire Kosher market is guaranteed to belost. If it is certified, then the company can compete for these customers.

When a product is certified Kosher, then it is Kosher for all of the market and not justsmall number of people who care about this issue. For the other 99% of the market, no

12 The GECAS orders would be adjusted to (83, 99, 110, 193, 393) for a total of 878 out of 15,979. Adjusting for the lower price ofregional jet engines, GECAS’s share would be below 5%.


45

harm is done. The small group of Kosher consumers is able to tip the market. Becausethey care and others are indifferent, manufacturers make choices based on thepreferences of less than 1% of the market.

Reynolds and Ordover suggest that GECAS’s favouritism towards GE products (andHoneywell post merger) will play a similar role. For GECAS, a plane is only “Kosher” if allof the parts that could be are, indeed, supplied by GE and Honeywell. Thus, an airframerknows that it would sacrifice all of GECAS’s business if it doesn’t equip the plane with GE/Honeywell parts where possible. If other customers don’t care, then GECAS’s preferenceswould determine the results for all of the market and competitors will be foreclosed.

The Commission asserted that competitors would be foreclosed because the airframer’spotential profits on lost airplane sales to GECAS will dwarf any discount or quality differencesthat Honeywell’s competitors could offer. This is because the profits on a $50 million B737plane far exceed the potential price or quality differences on a $100,000 piece of avionics.

Here some care must be taken. If the airframer is making the choice for its entire fleet,then the discount or quality difference should be multiplied by the expected sales. For a plane with sales of 1,000 units, even this small avionics part adds up to $100million, now comparable to the sale of two B737 planes.13 Of course, if GECAS were to order 50 planes and if all of those sales were to be incremental, then the scales would tip back the other way.

The incremental aspect is key. Unlike a sale to an airline, a sale to GECAS is a sale to a middleman who must then lease this plane to an airline. Coca Cola would lose sales to the Kosher population if their product were not certified. But if GECAS does not buythe plane, that does not mean the plane will not be sold. Had GECAS not made anyspeculative purchase of new aircraft, we do not believe that 1,192 fewer engines (andthe planes that go with them) would have been purchased during the period 1996-2000.Most (perhaps all) of these orders would have been picked up by ILFC and other leasingcompanies and by airlines purchasing directly from the airframers. Thus, we need toconsider the extent to which GECAS’s orders simply offset end-customer demand forplanes with GE/CFMI engines (and were the merger to go forward, Honeywell components).

Let us retrace the leverage argument a bit more slowly to see its limitations. This argumentprimarily applies to the situations in which the customer has no choice, i.e., the planeonly comes one way.14

For the most part, this limits the problem to what is called Supplier Furnished Equipment(SFE) as opposed to Buyer Furnished Equipment (BFE). An example of SFE would be theauxiliary power unit for a plane.15 In contrast, many avionics components such as cockpitdisplay systems are determined by the end customer from a choice of alternativesselected by the airframer.

13 Boeing 737 sales exceed 5,000. Sales of the 747 family exceed 1350 and the relatively new 777 family has sales of 606 as of 9/02.Data source is http://active.boeing.com/commercial/orders/customquery.cfm

14 Even where the customer does have a choice, the leverage argument suggests there is still a benefit to GE. GECAS’s favouritism willhelp ensure that GE is always one of those choices. But as that does not foreclose rivals, we continue our focus on the situationswhere the airframer makes an exclusive choice that determines the result for all customers in the market.

15 Even here there are exceptions. As noted by Reynolds and Ordover, on some A320 family planes, more than one APU has been certified.


46

There is no data that would reveal whether in its middleman role GECAS could forecloseHoneywell’s competitors in the SFE markets. The best we can do is look to see whateffect, if any, GECAS has had on shifting share in the engine market towards GE-poweredplanes. This is because prior to the merger, GECAS’s favouritism only extended to engines.

In the case of engines, most large commercial aircraft come with a choice of two orthree engine models. Thus an airframer does not make a choice that determines whichengine will go on the plane. There are some important exceptions to this rule, inparticular the case of regional jets, all of which come with only one engine choice.Reynolds and Ordover suggest that the regional jet manufacturers picked GE engines so as to be able to sell to GECAS. Later in this case, we examine the empirical evidenceand do not find it credible that the leverage theory played any role in the engine choiceof the regional jet manufacturers. Nor do we find convincing evidence of share-shifting in cases where customers did have a competitive choice of engines.

The Kosher certification model relies on the fact that a few customers care and the restare indifferent as to the certification. There is no one who opposes the certification. In the present case there are three sets of players who may prefer that the product isnot Kosher or GE/Honeywell-equipped. These three groups are

1. Rival engine and avionics/non-avionics supply companies

2. Airframers and airlines with a preference for non-GE/Honeywell parts

3. Other leasing companies that compete with GECAS

Just as Honeywell has an incentive for its components to be certified as SFE, competitorssuch as Hamilton Sundstrand have incentives to influence the airframer to make adifferent choice.

A food can have multiple certifications. It can be Organic, Kosher and Halal.16 One choiceneed not conflict with another. Each certification expands the customer base. By contrast,the airframer must choose either Honeywell or one of its competitors for each SFEcomponent. One choice excludes the other.

GE and Honeywell’s competitors have the ability to demonstrate the price/performancesuperiority of their products. They also have an ability to partner with a leasing companyor an airline. In either case, they could help create a customer with an anti-GE/Honeywellpreference.

Airlines themselves may have preferences for one manufacturer over another. This maybe due to performance differences or fleet standardisation issues. Thus an airframer thatstrongly favoured Honeywell parts would find its end product less attractive in the market.GECAS’s orders don’t help if the airlines are not ultimately attracted to the product.

Even in the speculative leasing market, there is reason to believe that GECAS’sfavouritism will be offset by its rivals. While it is natural to suppose that GECAS willfavour products containing GE parts, how would GECAS’s rivals feel about this issue?

16 Halal is an Islamic food certification.


47

In many industries, we observe that companies don’t like buying products that end upsupporting their rivals. This is an issue we have seen in many industries as a result ofour work developing and implementing Co-opetition (Nalebuff and Brandenburger, 1996).For example, in the quick-service restaurant industry, there was a period when Pepsiowned Kentucky Fried Chicken, Taco Bell, and Pizza Hut. Around that time, Burger Kingswitched from serving Pepsi to Coke. One of the factors influencing their decision wasthat buying Pepsi was supporting the parent of its rivals. Pepsi subsequently spun offtheir restaurant business and thereby eliminated this conflict of interest issue.

Similarly, AT&T felt the need to spin off Lucent. While Lucent was part of AT&T, many ofthe regional Bells were hesitant to purchase their switching equipment from someonethey saw as a rival.

Our examination of the data suggests there is some early indication of other leasingcompanies shifting away from GE-powered planes.17 Recall that GECAS is less than 1⁄3

of the speculative purchases of new aircraft by leasing firms. To the extent that the othertwo-thirds begin to favour non-GE equipped planes, the Reynolds and Ordover leverbegins to operate in the opposite direction.

Finally, we need to make a distinction between what GECAS might have been able toaccomplish with its GE-engine favouritism and what it might be able to accomplish wereit to extend this favouritism to Honeywell components. For the leveraging theory to work,

1. GECAS must extend its favouritism to Honeywell products.

2. Airframers must believe that adding more Honeywell components will lead toadditional GECAS sales.

3. Those additional sales to GECAS must not only be incremental for GECAS, they mustbe incremental to the airframer. The incremental sales to GECAS must not be offsetby any reduction in sales to airlines or to other leasing companies.

4. The profit on those additional airframe sales must be enough to offset the lower cost or higher quality of competitors’ components over the expected lifetime sales of the airframe.

We turn now to examine the logic and evidence for each of these steps.

5 Would GECAS extend its policy to Honeywell?

The European Commission assumes that:

Following the proposed merger, Honeywell’s [Supplier Furnished Equipment] productrange will benefit from ... GECAS’s instrumental leverage ability to foster the placementof GE products through the extension of its GE-only policy to Honeywell products.”18

Is this extension inevitable - or even likely?

17 Recall that GE’s participation in speculative leasing is relatively recent. Post acquisition of GPA, GECAS’s first order was in 1996. Thus any competitive reaction to GE’s connection to GECAS would be very recent. See section below on Competitive Reactions forOther Leasing Companies.

18 Paragraph 406 of European Commission Decision in Case No. COMP/M.2220.


48

The differences between Kosher certification and GECAS engine policy helps us understandwhy the GE-only engine policy may not extend to Honeywell’s numerous SFE components.Kosher is all or nothing. There is no such thing as being 99% Kosher. But for GE, a planewith a GE engine is already a valuable sale. In fact, recall that the majority of its enginesales are done through its joint venture with SNECMA. GECAS is happy to buy planesthat are desirable for its leasing business and are, in effect, only 50% powered by GE. There is no reason to think that post-merger GECAS would not want to buy apopular plane that has 50% GE power whatever the level of Honeywell components.

Would it be wise to give up the tens of millions of revenue on an engine sale and theprofits on the leasing business in order to obtain the tens or hundreds of thousandsrevenue on some piece of Honeywell avionics or non-avionics? It strains credibility thatGECAS would boycott a popular plane that came with a GE engine but did not haveenough Honeywell components.

To the extent one believes that GECAS has succeeded in shifting engine shares, this verybelief makes the potential cost/benefit analysis of extending the policy much lessattractive. The costs are higher because GE would be putting an engine sale at risk inaddition to GECAS’s leasing profits. The gains are lower because of the smaller dollaramounts for avionics and non-avionics parts.

When GECAS entered the market, most planes already came with a GE/CFMI engineoption, including the two most popular aircraft, the B737 and A320 families. But weknow of no existing planes that have all Honeywell components wherever Honeywelloffered a product. Thus GECAS would have to cease making speculative purchase of new aircraft if it adopted the same strict preference toward Honeywell componentsthat it has displayed for GE and CFMI engines.

If post-merger GECAS were willing to buy existing planes with 40% Honeywell componentsin some cases and 60% in others, why wouldn’t they also buy next-generation planesthat also came with some Honeywell and some of its competitors’ avionics and non-avionics components as supplier furnished equipment? Since airframers would expect tosell to GECAS in any case, there is no leverage.

But even if there is no leverage, why take a gamble? If the airframer expects a sale toGECAS, why forgo that sale over the choice of a small part? If the airframer is already99% Honeywell, why not go to 100% and be safe?

There are several answers. GECAS and GE have much more to lose than the airframer.GECAS is not the end customer, so losing a sale to GECAS does not mean the plane willnot be sold. If the airframer puts on inferior equipment or equipment at an inflated price,it will risk losing sales to end customers.

Secondly, the question starts with a false presumption. It is unrealistic to imagine thatthe airframer is considering going from 99% to 100%. Honeywell is unlikely to have thebest product or the best price performance for every SFE component. Thus there is areal cost in going from Honeywell’s historical market share to 100%. Given that 100% isan unobtainable goal, the question becomes: do we expect GECAS not to buy a plane


49

with GE or CFMI engines but only 47% Honeywell components and not 48%? Asking thisquestion reveals its folly. A distinctive feature of the engine choice is that it is onechoice, GE/CFMI or not. With avionics and non-avionics components, there are dozens ofchoices. Once the all-or-nothing aspect of the decision is lost, so is the leverage.

The most favourable but still plausible case for Reynolds and Ordover is that post mergerGECAS might exhibit some preference towards planes with more Honeywell products.19

This is wholly different than the claimed result with GE/CFMI engines. We have noHoneywell-only preference. Once GECAS is willing to buy a plane (because it has GE/CFMIpower), the airframer can expect sales to GECAS based on the plane’s potential forleasing profits no matter what the Honeywell share of equipment. The foreclosure theorywas based on the manufacturer’s fear that GECAS would “boycott” a non-GE/CFMIpowered plane. When applied to the Honeywell merger, the best theory of antitrustharm has come down to: GECAS, with its 5% to 10% market share might risk buying an extra plane or two if the airframer employs a few more Honeywell components.

Even this is not clear. The logic of Reynolds and Ordover — that the aircraft is so muchmore expensive than the individual components — cuts both ways. It makes theairframer keen to make an incremental sale. But it also makes taking on an incrementalaircraft very risky to GECAS because of the large potential losses if it cannot place theaircraft at a profitable lease (if it has taken on too many aircraft). Thus the smallincremental profits that accrue to GE for an extra $100,000 piece of avionics would bedominated by the concern to make a profit on the lease. The result would be little to no effect on its speculative purchases.

In fact, GECAS could even end up buying fewer planes. We have to assume that theHoneywell component does not offer the best price/performance deal as otherwise thecomponent would be chosen strictly on its merits.20 To the extent that the component’sprice or performance is inferior, this will reduce airlines’ demand for the plane. Even aslight reduction in demand on 95% of the market is not worth taking to get a slightincrease on 5% of the market. In fact, the inferior component may even reduce GECAS’sdemand - as GECAS has to lease the plane and thus its demand ultimately reflects theairline customer demand.

It should also be apparent that given the much reduced scale of potential incrementalGECAS purchases, it also becomes much easier for rivals to counterbalance this effect,should it arise. Inducing even one airline to express a preference for non-Honeywellcomponents would more than offset a small incremental GECAS order.

A small preference on a small market share of a market intermediary, all of which maybe counterbalanced by other players, does not provide a convincing theoretical basis formarket tipping, for market foreclosure, or for blocking a merger.

19 Reynolds and Ordover (2002) recognise this point in their footnote 43: “GECAS’s bias need not manifest itself in a 100% Honeywellpreference, but more embedded Honeywell equipment will increase GECAS’s aircraft preference in a continuous manner.” But theydon’t show how this continuous preference could lead to an Archimedean lever and market foreclosure.

20 The GECAS purchase might break a literal tie. On the other hand, price competition will leave few ties. If the airframer is trulyindifferent, this also provides an opportunity to send the order to a smaller supplier in order to preserve a competitive base ofsuppliers in the future.


50

In the empirical evidence section, we show that it is doubtful that even GECAS’s near100% preference for GE-engines has had any measurable impact on market shares.Note that this lack of a measurable impact is not on just a one-plane purchase - the sumtotal of GECAS’s purchases have had no demonstrated effect. This makes it even lesslikely that a small favouritism by GECAS for planes with Honeywell components wouldlead to a share-shift or, even more radically, foreclose its rivals from the market.

6 Reasons why GECAS has this policy

Underlying the entire concern about the role of GECAS is its favouritism policy. The original version of the theory supplied to the Commission by economists retained by United Technologies Corp. was premised on the assertion that this policy must becostly to GECAS and therefore it must have some hidden benefit.

At first, the cost seems obvious. Any restriction on one’s business means a potentiallysacrificed sale opportunity. Why forgo profits if there was not some other way in whichthis policy would redound to GE’s benefit?

That the policy has a net cost is subject to question. No attempt was made to measuremissed sale opportunities. Nor was serious consideration given to direct reasons whyGECAS might have such a policy in place. Thus, it is not clear on net that this policy wascostly, even without considering share-shifting.

A natural reason to favour internal purchases is the avoidance of double marginalisation.This is a standard explanation for vertical integration. Avoiding double marginalisation is anefficiency gain and should not be viewed anti-competitive. In light of GECAS’s small marketshare, it is implausible that competitors will be foreclosed or even harmed by this efficiency.21

Pursuing a specialised strategy provides a second motivation. At the Harvard BusinessSchool, one of Professor Michael Porter’s first lessons is that a strategy is defined bywhat you don’t do. A firm can’t do everything. Are there any advantages to specialisationfor GECAS?

One advantage is that the leasing companies will have greater product differentiation.Airlines will think of GECAS first when they desire a GE-powered plane. ILFC and othercompetitors will naturally shift towards Rolls Royce and Pratt & Whitney poweredplanes. This product differentiation will help all of the companies in this industry.

A very robust conclusion of economic theory is that there is no reason to expect that allfirms in a given market would systematically choose to offer the same product selection.In fact, there is an abundant literature documenting conditions under which companieswould definitely not want to adopt similar product mixes as leasing companies that “lookthe same” are likely to engage in fierce price competition leading to low profits.22

21 The issue of double marginalisation is subtle. If GECAS orders do not result in a share shift, then GECAS should include in the cost of an engine the full profits earned by GEAE. Double marginalisation is also less of an issue to the extent that contracts employnegotiated prices rather than fixed prices.

22 For an overview, see, Beath J. and Y. Katsoulacos, The Economic Theory of Product Differentiation, Cambridge University Press, 1991.


51

Specialisation is a matter of degree. Since GE/CFMI engines are an option on nearly allairframe options and are chosen by roughly half the time, this is a large market to goafter.23 GECAS also competes in the entire used market. It is very common to observecompanies entering one or more segments of the market and then expanding as theybecome successful. For example, the company might initially limit itself to the U.K. orGermany and then roll out to the rest of Europe. Or it might begin selling books and thenexpand to records. The fact that a company does not attack one hundred percent of themarket does not imply that the strategy has some hidden cost and therefore must havesome hidden benefit. Looking at GECAS’s growth over the period 1996—2000, it is hardto conclude that their performance was somehow hindered by having adopted a sub-optimal strategy.

A quite different explanation for the GECAS favouritism looks at behavioural factors, suchas company pride. Very few Ford employees drive to work in a Lexus. Those that do findthis is not the way to get promoted. Similarly, we understand a company that sends abusiness proposal to UPS using a FedEx carrier is unlikely to get the proposal read, nevermind win the business. Coke executives are known to leave business dinners if they aretaken to a restaurant that doesn’t serve Coke.

Buying a plane with a competitor’s engine would be difficult to explain. That should not leadus to conclude that the same policy would apply in the same way to Honeywell components.First, nearly all planes come with GE or CFMI engines as a choice. Thus, GECAS doesn’tfeel that it is excluded from any significant part of the market. In contrast, almost noplanes come with Honeywell parts for each and every component that Honeywellmakes. GECAS could not grow its leasing business unless it bought planes that were not100% Honeywell (where possible). Similarly, a Coke executive may choose to patronisea restaurant that serves Coke, but will consider other offerings when it comes to wine oreven bottled water (as Dasani and Evian have a much lower market presence).

A second observation is that aircraft engines are a flagship product for GE in a way thatfew other products are or could be. Because engines are expensive and the orders tendto be large, each competition matters. In contrast, Honeywell can win 42% or 63% ofthe multiple competitions for a given aircraft model. There are dozens of avionics andnon-avionics components that go into a plane. The lack of an all-or-nothing outcomewould make the Honeywell result less of a focal point.

The cost of GECAS favouritism can only be measured relative to the costs of alternativestrategies. Were GECAS to put rival engines on its planes, it would have to negotiatewith Rolls Royce or Pratt and Whitney over price. It is far from clear that GECAS wouldbe able to obtain the best prices on these engines. One reason is that rivals might beconcerned that their pricing information would get back to GE Aircraft Engines (GEAE).Rolls Royce and Pratt & Whitney would not want GEAE to know their lowest prices. Oneway to ensure that information does not get back is to not offer their lowest price to GECAS.

23 In fact, one could argue that 75% of the most relevant market is open to GECAS. The most popular narrow-body large commercialaircraft all have a CFMI engine. In this key arena for leasing companies, CFMI engines are chosen roughly 75% of the time, since theyare 100% of the B737s and about 50% of the A320 family. Indeed, of the total Boeing and Airbus aircraft ordered by GECAS morethan 83% (360 out of 433) have been in these two narrow-body aircraft families.


52

A second cost of employing rival engines is that competitors might use this purchase to create a public relations event. At Yale School of Management, we’ve experiencedsituations where Ford executives did not want to give a lecture in the GM room.24

GECAS would naturally be worried how rivals might exploit an opportunity where GEwas seen to have selected a rival’s engine over a GE engine. Does that mean that GE didn’t have the best engine?

To summarise, there are several reasons to suggest that GECAS’s favouritism is arational strategy on its own. We need not conclude that its hidden purpose must be to shift market share.

Nor can we conclude that this policy would have been extended to Honeywell postmerger. Even if one believes that GECAS has succeeded in shifting engine shares, thisvery belief makes the potential cost/benefit analysis much less attractive. The costs arehigher because GE would be putting an engine sale at risk in addition to GECAS’s leasingprofits. The gains are lower because of the smaller dollar amounts for avionics and non-avionics parts. Finally, since the Honeywell components decisions will not be 0% or 100%,it is much less convincing to build a lever off a shift from 47% to 48% Honeywell parts.

7 Empirical support

FACTS AS THEY APPLY TO REGIONAL JETS

In an attempt to demonstrate the Archimedean lever, the European Commission lookedat the case of the engine choice in the large regional jet market. Those citing evidence ofa market tipping observe two facts.

1. Three regional jet manufactures each chose a GE engine as their exclusive power source.

2. GECAS subsequently ordered 150 regional jets, 50 from each of Bombardier,Embraer, and Fairchild Dornier in 2000 (with options on another 100 jets from each).

These two facts led some people to conclude that there was some either explicit orimplicit expectation that by choosing the GE engine, an order from GECAS would follow.For example, Reynolds and Ordover write:

Given GECAS’s position by the late 1990s as a purchaser of large commercial aircraftand in light of subsequent events, it would be surprising if none of the regional jetairframers anticipated such orders.

While this fact pattern is consistent with the tipping model, it leaves out a critical timingdetail. As Reynolds and Ordover themselves recognise, at the time all three enginechoices were made, neither GECAS nor any other leasing company had made anyspeculative purchases of a regional jet.25 In fact, the Bombardier engine choice on the

24 This created a problem as the GM room (renovation paid for by GM) is the single largest lecture space at SOM. While there is FordCourtyard, this outdoor space is not suitable for a lecture.

25 See Reynolds and Ordover points (d) and (e), page 188. The small order by Bavaria International Leasing Company was in June, 2000,after the engine choices were made.


53

CRJ-700 was made in 1994, well prior to GECAS having made any speculative purchaseof any aircraft. (The Fairchild Dornier engine decision was made in 1998 and the Embraerengine decision was made in June 1999.)

According to GE, it was only after the engine decisions of Bombardier, Embraer, andFairchild Dornier were made that GECAS received the results of two managementconsulting studies evaluating the leasing market for large regional jets. The reports came back positive and shortly thereafter GECAS contacted the three regional jetmanufacturers to explore the purchase of their large regional jets.

It strains credibility to presume that a regional jet manufacturer would favour GE engineson the hope that GECAS would later enter a market that no one had previously enteredand then purchase its regional jets.26 Of course, if such expectations had existed, we would expect to have heard them recounted. No such evidence was presented. To the contrary, evidence was presented that GECAS did not influence the engine selectionprocess for the large regional jets.

While the European Union and the United States have different antitrust rules and thereforereach different conclusions based on the same fact pattern, they should reach the sameview of the facts. According to Deputy Assistant Attorney General William Kolasky,

We examined each of the four transactions in which GECAS allegedly used its buyingpower to get GE sole source on new aircraft platforms. In each, we found thatGECAS played no role in the customer’s decision to sole source and that in each caseGE had won the competition on the merits, by offering the best engines for thecustomer needs at the best price.27

This is in line with GE’s explanation for why all three regional jet airframers chose GE power.28 The first entrant, Bombardier, had a long-standing relationship with GE. No existing engine was suited for the large regional jet. GE took a gamble and agreed tobuild a larger model in the CF34 engine family, thereby allowing Bombardier to developthe market with their CRJ-700. The large regional jet market turned out to be a biggermarket than anyone anticipated. Fairchild Dornier and then Embraer each looked to enterthe market quickly. GE had a real engine; competitors’ engines were all on the drawingboard. To speed their entry, both of these airframers chose GE.

The regional jet market was the European Commission’s best example of a tipping theoryin action. The circumstantial evidence seems impressive until one takes a closer look.

PLANE WITH COMPETITIVE ENGINE CHOICES

Even as we have narrowed our focus to the speculative leasing part of the business, forsome of our purposes this is still too broad. The issue at hand is whether GECAS couldplay a role in shifting the market towards Honeywell components in the event of amerger. While there are many important differences between Honeywell’s components

26 To which Reynolds and Ordover might reply: that shows just how strong GECAS’s power is! Thus we now have a theory where apotential entrant with zero percent market share can tip the market and thereby foreclose rivals.

27 William Kolasky, “Conglomerate Mergers and Range Effects: It’s a Long Way from Chicago to Brussels,” Address given before GeorgeMason University Symposium, 11/9/01, pages 18-19. This also rejects the view that GECAS was a “launch” customer.

28 Note that GE has not won all recent engine competitions for regional jets. Embraer selected Rolls Royce over GE to supply the engineon the EMB 145 and then picked GE for the EMB 170/190.


54

and engines, as a first cut, it is instructive to see whether GECAS has been able to playany role in shifting engine sales towards GE power prior to the merger. In some sense,this is a necessary but not sufficient condition. If GECAS had not been able to influencethe engine market, then it was even less likely that the merger would somehowinfluence the avionics and non-avionics components market.

The place to look for evidence of share-shifting are the cases where there is a choice ofengines. For example, the 737 plane only comes with a CFMI engine.29 In that sense,GECAS, by its choice of engine can not influence market share, as anyone who buys thisplane will automatically get a CFMI engine. In other cases, such as the Boeing 757,customers do have an engine choice, but GE power is not among them. (Rolls Royce andPratt and Whitney are the two power choices.) Here again, we do not look for evidenceof GECAS’s favouritism towards GE power having a share-shifting effect. Thus, in this casestudy, we will often focus on what will be called the “competitive engine” choice market.For the planes in this market, there is a GE power option and at least one other choice.

FIGURE 2

SIZE OF SPECULATIVE LEASING MARKET

FOR PLANES WITH COMPETITIVE ENGINE CHOICE

Figure 2 above shows the size of the speculative leasing market for planes with acompetitive engine choice.

We can get a better understanding of GECAS’s role in the market by looking at theengines ordered by GECAS and its rivals.

Total Engines

0

100

200

300

400

500

600

700

91 92 93 94 95 96 97 98 99 2000

29 This decision by Boeing was made in 1993 — before GECAS began making speculative purchases.


55

The total orders by leasing companies are divided into three groups. The bottomsegment are the GECAS purchases. The next segment are purchases of GE/CFMIengines by rival leasing companies. The top segment are purchases of rival engines byother leasing companies.

In this chart, one can observe that GECAS made large purchases in 1996 in order tobuild up inventories. Over the following three years, the majority of rival leasing companychoices were not GE or CFMI. In particular, 1999 was a year where rivals placed a largenumber of orders and chose relatively few GE/CFMI engines. Note that the lowernumber of engines in 2000 compared to Figure 2 reflects the fact that for mostpurchases (other than those by GECAS) the engine has not been specified. Only a littleover 200 of the more than 600 engines were determined as of the time of this date, 6 May 2001. Thus the year 2000 column suggests a bigger role for GECAS in that all ofits engine choices are made at the time of order.

FIGURE 3

SPECULATIVE LEASING COMPETITIVE ENGINE CHOICE

EMPIRICAL EVIDENCE OF SHARE SHIFTING

We turn now to consider the central empirical question in this case: was GECAS able toshift share in favour of GE/CFMI engines? We ask this question both for its own sake andas a way to get insight into whether GECAS would be able to shift share towards Honeywell.

Several commentators on this issue (Kolasky 2001, Pflanz and Caffarra 2002) haveemphasised the need to determine whether GECAS’s orders would have an effect in

0

50

100

150

200

250

300

350

400

91 92 93 94 95 96 97 98 99 2000Year

Num

ber o

f Eng

ines

Non-GECAS rival engine purchases

Non-GECAS GE/CFMI purchases

GECAS GE/CFMI purchases


56

equilibrium, that is, after we take into account the reaction of other players. GECAS’sfavouritism could be offset by other leasing companies. Their purchases could also havedisplaced direct airline purchases?

Reynolds and Ordover were aware of this issue and tried to provide evidence of GECAS’sability to create an equilibrium market share shift. They base their conclusions on thefollowing data.30 Prior to GECAS’s entry in 1996, GE/CFMI’s share on planes with acompetitive engine choice was just over 50% and that share fell by 5% to 45% in the 1996-2000 GECAS period during which GECAS has placed speculative orders for new aircraft.

In contrast, prior to GECAS’s entry, speculative leasing companies chose just over 40%GE/CFMI engines and after GECAS’s entry, that percentage rose to just over 62%.Putting these two effects together, they claim the leasing company mix was 27% out of balance in that it rose by 22% at a time when the GE/CFMI’s share of airline orderswas falling by 5%.

This analysis of Reynolds and Ordover is echoed almost exactly by the EuropeanCommission case decision:

GECAS has indeed been able to significantly increase GE’s position without GECAS’sincreased purchases of GE engines having been offset by purchases of non-GEengines by airlines and its other leasing companies. Consequently, through GECAS’sbias in favour of GE engines and its influence over airlines, GE has been able toincrease GE’s market shares of engines.31

A comparison of GE’s market position pre-GECAS (from 1988 to 1995) with the post-GECAS situation (1996 to 2000) shows that while GE’s engine sales with leasingcompanies, including GECAS, increased by 20 share points (or over 60%), the directpurchases of GE engines by the airlines only dropped by less than 5 share points (or less than 10%). The fact that other leasing companies and airlines simply have notcompensated for GECAS’s biased purchases results in a net shift of engine marketshares to GE.32

The reader might be surprised to learn that comparing the 1991-1995 pre-GECAS periodto the 1996-2000 period, GE/CFMI’s overall share of the competitive engine choice marketwent down - from 52.3% to 50.6%.33

How is that possible? The problem is that the European Commission and Reynolds andOrdover facts do not prove that a share shift occurred. Nor are they even suggestive. A simple example will explain why. Assume as a starting point that airlines buy planes with

30 See Figures 2 and 3 of Reynolds and Ordover.

31 Paragraph 136 of European Commission Decision in Case No. COMP/M.2220

32 Paragraph 138 of European Commission Decision in Case No. COMP/M.2220.

33 The competitive engine choice market is the market for engines on all planes where the airline has a choice of two or more engines toput on a plane and where one of those choices is GE/CFMI. In the 1991-1995 period, GE/CFMI engine orders on competitive planestotalled (1,001+160) out of (1,917+300) for a market share of 52.3%. (The first number is airline purchases and the second number isleasing purchases.) In the 1996-2000 period, GE/CFMI engine orders on competitive planes totalled (2,188+778) out of (4,656+1,208)for a market share of 50.6%. Just as noted by the MTF, GE’s share of the airline market fell by five share points, from 1,001/1,917 or52.2% to 2,188/4,656 or 47%. And their share of the leasing market rose from 160/300-or 53.3% to 778/1,208 or 64.4%. Note thatthe rise is less than the number used by the MTF and Reynolds and Ordover as their pre-GECAS period extends back to 1988. But thisdifference is not material to the conclusion. The important point is that if GECAS shifted share from direct purchases to leasing, wewould expect to see exactly what was found in the data with essentially overall share shift.


57

100 engines, 50 GE and 50 others. There is a leasing business which buys planes with10 engines, 5 GE and 5 others. Now GECAS enters and buys planes with 5 GE engines.Because it leases these planes to an airline, that airline no longer purchases the planedirectly. Thus the new result is that airlines buy planes with 95 engines, 45 GE and 50others. Leasing companies buy planes with 15 engines, 10 GE and 5 others.

Nothing has changed in terms of GE’s overall share. They had 50% share before andthey have a 50% share afterwards. But their share in the airline business has gone downby 5% while their share in the leasing business has gone up substantially, from 5 out of10 to 10 out of 15 (50% to 66.7%). Using the logic of the MTF, one would conclude thatGECAS had shifted share due to their 21.7% shift (leasing went up by 16.7% whileairlines went down by 5%).

The false conclusion results from the fact that leasing is small relative to airlines andtherefore the two percentages cannot be added. Second, GECAS orders can substitutefor airline orders, not just other leasing companies.

During the period in question, the speculative leasing business tripled in size. Half of thatgrowth can be attributed to the emergence of GECAS. One reason why we expect tosee the GE engine share of airlines go down and the GE engine share in leasing go up isthat some airline customers who were looking to buy GE powered planes shifted toGECAS. Thus the use of GE’s engine share of airline purchases is not a valid baseline.

Our numerical example above was simplified for clarity. We now develop this examplemore carefully using actual market numbers and allowing for substitution betweenleasing companies.

In the period 1991-1995, airlines ordered 1,917 engines on planes with a competitiveengine choice.34 The next five-year period, 1996-2000, airlines ordered 4,656 engines on planes with a competitive engine choice.

The speculative leasing companies, including GECAS, increased the size of their ordersfrom 300 to 1,814, between these two periods. GECAS accounted for 488 of those 1,814.

MARKET SHARES OF AIRCRAFT ENGINES

1991-1995 1996-2000

Airlines 86.5% 72%

Non-GECAS 13.5% 20.5%

GECAS 0% 7.5%

34 Reynolds and Ordover use 1988-1995 as the pre-GECAS period. Our dataset is the same, but it does not include the years 1988-1990.However, our results for 1991-1995 are very similar to those of Reynolds and Ordover found for 1988-1995 and thus there is no reasonto suspect that this difference has any bearing on our different conclusions.


58

Consider the hypothesis that GECAS has had no influence on GE engines’ market share.If so, then GECAS 7.5% of engines is simply going to airlines that otherwise would haveordered GE-powered planes, either directly from the airlines or from other leasingcompanies. Is this consistent with the data?

Take GECAS’s 7.5% share and divide it up between airline purchases and other leasingcompanies in the same proportion as their market shares (72:20.5). The result is ourright-hand column.35

EFFECT OF REMOVING GECAS FROM ENGINE DECISION

With GECAS Assuming no GECAS

Airlines 72 total engines 78 total engines

32.5 GE engines 38.5 GE engines

39.5 other engines 39.5 other engines

GECAS 7.5 total engines 0 total engines

7.5 GE engines 0 GE engines

0 other engines 0 other engines

Other Lessors 20.5 total engines 22 total engines

8 GE engines 9.5 GE engines

12.5 other engines 12.5 other engines

The new mix of airlines is almost 50-50 (it is 49.4 to 50.6). This is roughly the same mixas the airlines had prior to GECAS’s entry in 1996. As for other leasing companies, themix is 40:60. This, too, is almost same as prior to GECAS’s entry in 1996.

Looking at the same data from this perspective, we conclude that the data is entirelyconsistent with the conclusion that GECAS has had zero impact on shifting the marketshare of GE engines.

The key difference between the two approaches is that Reynolds and Ordover implicitlyassume that none of GECAS’s market share came from a reduction in airline purchases.We take the view that as leasing companies are only market intermediaries, the bulk ofthe GECAS share will come out of direct airline purchases. Evidence for this is the factthat the leasing business grew much faster than the total airline business.

35 Note that we use a base of 100 engines so that engines represent market share (which explains the fractional engines).


59

In the case of competitive engine choice, the leasing companies’ market share grewfrom 13.5% to 28%. A similar growth was also seen in the overall leasing market.Between 1991-1996, the leasing companies had a market share of 11.7%, which grewto 23% in 1996-2000. The yearly changes can be seen in the chart below with thebottom segment representing GECAS, the middle segment other leasing companies,and the top segment airline purchases.

FIGURE 4

GECAS/RIVALS/AIRLINES TOTALS

When GECAS orders a large number of planes with GE engines, it will follow that someof the airlines will be buying fewer of those planes directly (and obtaining more of thoseplanes from GECAS). One can not use the market share of GE engines among airlinespurchases as a measure of market preference of comparison as this market share isendogenous to GECAS’s size in the market.

How then would one show an equilibrium share shift? One way would be to demonstratethat the total percentage of GE power rose during the period in question. That is, the largenumber of GE/CFMI engines purchased by GECAS was more than the combined reductionin GE/CFMI purchases by the airlines and other leasing companies. The data show thatdid not happen. GE’s overall market share in the competitive engine purchases fellslightly between the 1991-1996 period and the 1996-2000 period, from 52.3% to 50.6%.

A re-examination of the same data does not lead us to conclude that GECAS had anymarket share shifting influence. The fact that we reach a different conclusion thanReynolds and Ordover while using the same data does not mean that the answer simply

0

500

1000

1500

2000

2500

3000

3500

4000

4500

91 92 93 94 95 96 97 98 99 2000Year

Engi

nes

Airlines

Non-GECAS Leasing

GECAS


60

depends on how one looks at the problem. In order to believe that the GECAS entry intothe speculative purchase of new aircraft has led to a market shift in favour of GEengines, one would have to argue that GECAS did not expand the leasing market, butrather primarily took share from leasing rivals. This assumption is not supported by thedata. In contrast, the assumption that GECAS’s activity primarily grew the leasing market leads to GE/CFMI engine market shares among airlines and other leasing companiesthat are almost identical to the market shares prior to GECAS’s entry.

COMPETITIVE REACTIONS FROM OTHER LEASING COMPANIES

In response to GECAS’s entry, we might expect other leasing companies to respond byshifting their speculative purchases away from GE power. This is for several reasons asdescribed earlier in the section “Why GECAS has this Policy.” One reason is productdifferentiation. A second is that to the extent that GECAS has made orders to supplysome of the customers who desire GE power, supply and demand considerations willlead GECAS’s competitors to favour supplying airline customers with preferences forRolls Royce or Pratt and Whitney engines.

Reynolds and Ordover look for this effect by comparing the 1988-1995 period to the post GECAS 1996-2000 period. They find no large shift away from GE/CFMI engines in the aggregate data. But a closer look at year-by-year orders suggests that as GECAShad grown, there may have been a response from other leasing companies.

Consider the orders of ILFC, GECAS’s largest rival and a firm with roughly 1⁄3 of thismarket. In 1996 and 1997, 150 out of the 260 competitive choice engines wereGE/CFMI. For the latter three years, 1998-2000, only 26 out of the 104 engines chosenwere GE/CFMI. At the time of this data set (May 6, 2001), most of the recent ILFCorders left the engine choice unspecified and so this conclusion is still speculative. Still, it is suggestive that there has been a move away from GE power.

Additional evidence was seen in the previous Figure 3, which suggested that followingGECAS’ initial orders in 1996, the majority of rival leasing company choices shifted awayfrom GE or CFMI.

8 Should we be concerned?

To conclude this case study, we ask whether there should be a concern even if GECASdid have an ability to shift share.

If other manufacturers can offset this influence, there is no cause to take action. Herewe make two observations. Whatever barriers to entry exist, following the purchase ofthe bankrupt Guinness Peat Aviation, GECAS has grown to be one of the two largestspeculative leasing companies in a span of seven years. Second, Rolls Royce has a 50%interest in Pembroke, a small aircraft leasing company.


61

The MTF did not believe that others could enter and participate at the same level asGECAS.36 But if the Archimedean lever is correct, even participation at a small scale canhelp tip back the market, especially if the purchases are focused on cases where themanufacturer is making an all-or-nothing decision. Even without buying a leasing company,a manufacturer can create an alliance with an existing company, which would give thatcompany a preferred position. Again, if the European Commission is right that endcustomers have little preference for components so long as they meet quality standards,ILFC and others should require little compensation to favour non-Honeywell components.

9 Conclusions

Antitrust authorities are familiar with the complaint that a customer has sought toleverage its size so as to distort competition in favour of its related businesses. A clearexample is when Toys R Us attempted to leverage its 20% market share in the retail toybusiness by prohibiting manufacturers from selling products to warehouse club storesthat they also sold to Toys R Us.37 Manufacturers felt pressured to accept this demandfrom Toys R Us, although it would be against their interest.

The case of GECAS does not fall into a traditional leverage theory. Their market share ofplanes or engines (measured in dollars) is in the single digits. They are not an endcustomer themselves - rather, they take possession of planes and attempt to lease themto an end customer. The cause for concern was that GECAS has demonstratedfavouritism towards GE engines. Our analysis in this case study suggests that this policyhas had no demonstrated effect on the market. Furthermore, this favouritism is unlikelyto be extended to Honeywell components, at least not in an all-or-nothing form. As wewould expect GECAS to continue buying planes with a mixture of Honeywell and non-Honeywell components, this suggests that the proposed merger would not create aleverage opportunity for GECAS.

The Commission’s focus on the role of GECAS is puzzling. One can interpret from itsactions that the European Commission objected to GECAS’s policy of favouring GEengines, separate from its implications for the merger. The Commission’s attemptedremedies seemed to go beyond protecting Honeywell’s competitors from foreclosure to also addressing competition in the engine market. A commitment not to favourHoneywell components would have prevented the Archimedean lever from gettingstarted. But the Commission did not accept such a commitment. Its preference againstbehavioural remedies and for structural remedies led them to push for divestment. GE was unwilling to sell off a substantial interest in one of its core businesses (to a thirdparty approved by the Commission).

36 We should be careful to distinguish between the ability to enter and the desire to enter. There are two interpretations to the fact thatUnited Technologies Corporation and other component manufacturers have not copied the GE strategy of owning a leasing company.One view is that entry is too difficult or expensive. A second view is that there are no share shifting gains and hence no benefits fromentering the speculative leasing business.

37 This practice was challenged on several grounds, one of which was that the manufacturers, in effect, had made an illegal horizontalagreement to boycott warehouse club stores. See http://www.ftc.gov/os/1997/ 9709/index.htm#30 and Toys “R” Us, Inc. v. FTC, No.98-4107, 221 F.3d 928; 2000 U.S. App.


62

Even a year later, one is still taken aback by the aggressive position of the EuropeanCommission in this case. This is a vertical foreclosure case against an incumbent with a5-10% market share. We have an untested theory that is intriguing but breaks downwhen applied to the specifics of Honeywell’s business. We have a case study analysis ofregional jet purchases and an interpretation of share-shifting data that were both flawed.

In spite of the Commission’s efforts to develop new theories and its carrying out anextensive examination of the market, we are left with no real understanding or evidenceof how conglomerate power is exercised. Or, if it exists at all.

There was a time when antitrust authorities had a “big is bad” view. GE is clearly big.While some practitioners may still have a gut instinct that large firms are able to leveragetheir power in and across markets, the GECAS aspect of the GE/Honeywell merger case demonstrates that this instinct leads to ungrounded decisions and ad hoc policy.38

In short, this is a case study of how not to conduct an antitrust analysis.

We take the view that caution is called for when considering conglomerate effects.Starting with Proctor and Gamble/Clorox, history has not been kind to the decisions thatblock conglomerate mergers.

38 See Patterson and Shapiro (2001). They argue that the GE-Honeywell case suggests the need for a better check and balance systemon the power of the European Commission to block a merger. According to a recent New York Times article, Commissioner Montiseemed to agree: “Mr. Monti also plans to propose naming a chief economist to oversee major merger assessments, and to use adevil’s advocate system in which teams of officials in his office search for weaknesses in their colleagues’ work before a decision ispublished.” New York Times, October 26, 2002.


63

VIIndependent ServiceOrganisations v KodakIn 1987, a number of Independent Service Organisations (ISO) sued Kodak over its policyof refusing to sell its parts for Kodak copiers directly to ISOs. Kodak was willing to sellreplacement parts to end-users who could service the copiers themselves and Kodakwas also willing to service the copiers itself, but it was not willing to sell the parts to the ISOs who would then provide the maintenance on the copiers. This case made it to the US Supreme Court which ultimately found that Kodak had violated the antitrustlaws. We will examine the facts of the case (as we know them) to try and determinewhy Kodak would refuse to sell its parts to ISOs. We conclude that the policy was a form of bundling that ultimately could lead to consumer harm in the short run as wellas in the long run.

1 Background1

The Eastman Kodak Company, which is best know for its photography products, alsomanufactures, sells and provides service for high-volume photocopiers and micrographicequipment. The specific photocopiers and duplicators involved in the case were thosemanufactured by Kodak under the trade name “Ektaprint.” As mentioned above, it wasKodak’s parts policy which led to the filed lawsuit.

The court in this case defined two distinct product markets for the original equipment:high-volume photocopiers and micrographic equipment. The ‘high volume photocopier’market included the Ektaprint family of copiers that range in copy speeds from 30 pagesper minute (ppm) to 100 ppm. Dataquest classifies mid volume copiers as those thatmake 30 to 69 ppm and high volume copiers as machines with the ability to make 70 copies or more per minute. Table 1 provides information on the sales of mid-volumecopiers from 1986 to 1991. Table 2 present the same information for high-volume copiers.

1 The information in this section comes from the decisions in the three courts that heard this case. Northern District of California Court,Kodak, (1988) WL 156332, the Ninth Circuit Court of Appeal (1990) 903 F.2d 612, and the Supreme Court (1992) 504 U.S. 451.


64

TABLE 1

MID VOLUME COPIERS (MARKET SHARES % BASED ON SALES)

1986 1987 1988 1989 1990 1991

Xerox 37.68 37.93 37.24 37.61 37.79 37.80

Kodak 4.89 4.55 4.91 5.00 5.33 5.56

IBM 1.90 1.90 1.48 0.54 0.05 0.00

Canon 8.42 9.99 10.30 10.38 10.50 10.61

Konica 5.09 4.17 4.20 4.18 4.22 4.21

Ricoh 4.41 5.50 5.68 5.60 5.68 5.70

Savin 11.27 8.91 8.70 8.42 8.09 7.90

Sharp 9.50 8.09 7.82 7.83 7.79 7.80

Pitney 2.72 2.97 3.32 3.53 3.72 3.88

Other 14.12 15.99 16.34 16.90 16.83 16.54

Source: Dataquest estimates for 1986 and 1987 company-specific data and for 1986-1991 total placements. Dataquest placement estimates are shipments less rental retirements making actual unit sales higher for most companies. Prudential-Bache Securities estimates are used for 1988-1992 company-specific placements.

TABLE 2

HIGH-END COPIERS (MARKET SHARES % BASED ON SALES)

1986 1987 1988 1989 1990 1991

Xerox 50.00 51.33 50.81 53.65 54.05 54.20

Kodak 21.01 27.67 27.04 34.60 35.14 35.10

IBM 20.17 11.00 10.75 0.95 0.00 0.00

Canon 5.88 5.67 5.54 5.71 5.70 5.40

Other 2.94 4.33 5.86 5.08 5.20 5.30

Source: Dataquest estimates are shipments less rental retirements. Actual unit sales are higher for most companies. Further, rental retirements are the largest in the high-end and mid-range sectors.

These market shares tables demonstrate that there were several significant competitorsin these markets. Kodak’s high volume photocopier products faced competition fromXerox, IBM, and Canon, while its competitors in the mid volume market include thesethree as well as Minolta, Bell & Howell, and 3M. Competition in the aftermarket (that isthe market for spare parts and service), though, was different than in the foremarket (the


65

market for the copier). Kodak’s parts were not interchangeable with parts used in othermanufacturers’ equipment and vice versa, so once a consumer had purchased a copier,there were more limited opportunities for spare parts.

In addition to selling the equipment, Kodak sold and installed replacement parts for itsequipment. The company dominated the service market for its own products until theearly 1980s, providing service for at least eighty percent of the machines itmanufactured. In selling the copiers and service, Kodak chose not to sell a completesystem of original equipment, lifetime service, and lifetime parts for a single price.Instead, Kodak provided service after the initial warranty period either through annualservice contracts, which include all necessary parts, or on a per-call basis. The particularprice it charged for equipment, service, and parts was different for different customers.

Kodak had ready access to all parts necessary for repair services because itmanufactured many of the parts used in its equipment and purchased the remainingnecessary parts from independent original-equipment manufacturers (OEMs). Kodak held220 United States patents covering 65 parts for its high volume photocopiers andmicrographic equipment. All Kodak diagnostic software and service software werecopyrighted or patented. For those parts on which it did not hold a patent, Kodak hadcontractual arrangements with the OEMs of these parts.2

Independent service operators began servicing Kodak equipment in the early 1980’s.ISOs also sold parts and reconditioned Kodak equipment. In general, the ISOs providedservice at a price substantially lower than Kodak.3 Some customers found that the ISOservice was of higher quality. In some cases, ISOs’ customers purchased parts fromKodak and hired the ISO only for service. Others chose ISOs to supply both service andparts. In order to be able to provide this service, ISOs would keep an inventory of parts,purchased from Kodak or other sources. In addition to the OEM’s, other sources ofKodak parts include: a) brokers who would buy parts from Kodak, or strip used Kodakequipment to obtain the useful parts and resell them, b) customers who would buy partsfrom Kodak and made them available to ISO’s, and c) used equipment stripped for parts.

In 1985, Kodak stopped selling its photocopier parts for new model machines to ISOs.Kodak adopted the same policy for its micrographic parts in 1986. Kodak sought to limitISOs access to other sources of Kodak parts through contractual agreements with itsOEMs. Kodak and the OEM’s agreed that the OEM’s would not sell parts that fit Kodakequipment to anyone other than Kodak. Kodak also pressured Kodak equipment ownersand independent parts distributors not to sell Kodak parts to ISO’s. In addition, Kodaktook steps to restrict the availability of used machines. Kodak intended, through thesepolicies, to make it more difficult for ISOs to sell service for Kodak machines.

2 The Court of Appeal found evidence that Kodak controlled a monopoly share in the market for Kodak parts through: “(1) its ownmanufacture of Kodak parts (30%); (2) its control of original-equipment manufacturers’ sale of Kodak parts to ISOs through toolingclauses (20-25%), engineering clauses and other proprietary arrangements; and (3) its discouragement of self-servicing and resale ofparts by end users.” Eastman Kodak Company v. Image Technical Services, Inc., et al. the United States Court of Appeals for the NinthCircuit no. 90-1029, page 11445.

3 See, e.g., Trial Exhibit 729, Image Technical Service, Inc., et al. v. Eastman Kodak Co., C 87-1686 (January 18, 1996).


66

It succeeded. ISOs were unable to obtain parts from reliable sources and many wereforced out of business, while others lost substantial revenue.4 Customers were forced toswitch to Kodak service even though they preferred ISOs service.

Several ISOs brought a lawsuit against Kodak in 1987 for violations of the Sherman Act.The ISOs’ principal allegations were that Kodak had unlawfully tied the sale of service forKodak machines to the sale of parts, which was in violation of Section 1 of the ShermanAct and that Kodak had monopolised or attempted to monopolise the sale of service forKodak machines, which was in violation of Section 2 of the Sherman Act.

The District Court of the Northern District of California granted a summary judgment infavour of Kodak. As to the first claim, the court found that respondents had provided noevidence of a tying arrangement between Kodak equipment and service or parts. Thecourt, however, did not address respondents’ first claim on a tying arrangement betweenKodak parts and service and between Kodak equipment and service. As to the secondclaim, the District Court concluded that, although Kodak had a “natural monopoly overthe market for parts it sells under its name,”5 a unilateral refusal to sell those parts toISOs did not violate Section 2.

The Court of Appeal for the Ninth Circuit reversed this decision. The Court, disagreeingwith Kodak, recognised that there are separate markets for service and parts. For serviceand parts to be considered two distinct products, there must be sufficient consumerdemand so that it is efficient for a firm to provide service separately from parts. Evidenceon the record indicates that service and parts have been sold separately in the past andstill are sold separately to self-service equipment owners.6 The Court also found sufficientevidence of a tie between parts and services. Furthermore, it was deemed that Kodak hasmore than sufficient power in the parts market to force unwanted purchases in the tiedmarket (service). Finally, the Supreme Court affirmed the decision of the Court of Appealagainst Kodak. We discuss both Kodak’s position and the Supreme Court decision below.

Kodak’s core defence focused on the existence of competition in the equipment market.The central argument was that any increase in profits from a higher price in theaftermarkets (i.e. parts and service), would be offset by lower profit in original equipmentsales. Thus, a large or even monopoly share of the relevant parts and service saleswould have not allowed Kodak to raise the overall price of copier services (whichincludes the original equipment, parts, and service) above the level that would becharged in a competitive market. In other words, equipment competition precludes theability to exercise monopoly power in derivative aftermarkets.

With respect to the second claim on the use of monopoly power to foreclosecompetition to gain a competitive advantage or to destroy a competitor, the companyalso provided business reasons to justify its conduct. Kodak contended that it had threevalid business reasons for its behaviour:

4 While often an ISO can conduct a repair without using a part that has a Kodak patent, the only way the ISO can guarantee the ability toconduct any service for a customer is if the ISO has access to all parts for the copier, including those parts under a Kodak patent.

5 See North District of California Court, Kodak, (1988) WL 156332.

6 At least some consumers would purchase service without parts, because some service does not require parts, and some consumers,those who self-service for example, would purchase parts without service.


67

1. To foster interbrand equipment competition by allowing Kodak to stress the quality of its service.

In defence of its anticompetitive conduct, Kodak contended that quality controlwas a legitimate business reason for its conduct. Kodak declared that, by preventingcustomers from using ISO’s, “it [could] best maintain high quality service for itssophisticated equipment” and avoid being “blamed for an equipment malfunction,even if the problem is the result of improper diagnosis, maintenance or repair byan ISO.”7

2. To improve asset management reducing inventory costs.

Kodak claimed that it also needed to be the sole service provider so that it couldcontrol its parts inventory costs.

3. And to prevent ISOs from free riding its capital investment in equipments, parts and service.

Kodak claimed that its policies prevented ISO’s from “exploiting the investmentKodak has made in product development, manufacturing and equipment sales inorder to take away Kodak’s service revenues.”8

The Supreme Court, based on the available evidence dismissed these arguments. The first argument put forward by Kodak, that its behaviour was necessary to maintainthe quality of the product, is a standard defence in tying cases. The court ruled that theevidence of ISOs providing quality service at lower prices was sufficient to refuteKodak’s argument.9 As a side consideration, the Court mentioned that if that was indeedthe case, then Kodak might have been willing to take advantages of low cost services toexpand its equipment base.10 The court also raised the issue of Kodak making conflictingstatements about customers’ informational sophistication. On one hand, Kodak claimedthe exclusive-service contracts were warranted because customers would otherwiseblame Kodak equipment for breakdowns resulting from inferior ISO service. On the otherhand, Kodak claimed that its customers would be able to fully evaluate the life-cyclecosts of buying, using, and maintaining its copiers. Kodak failed to provide reasons whyits customers should be able to perform complex lifecycle-pricing decisions and yet beunable to distinguish which breakdowns are due to bad equipment and which are due tobad service. In addition, because self-service customers were just as likely as others toblame Kodak equipment for breakdowns resulting from (their own) inferior service, Kodak’swillingness to allow self-service casts doubt on its quality claim. The Supreme Courtagreed then with the Court of Appeal that respondents “have presented evidence fromwhich a reasonable trier of fact could conclude that Kodak’s first reason is pretextual.”11

7 504 U.S. 451, Section IIIB. A similar claim was made, and rejected, in the IBM antitrust case. IBM argued, in the absence of evidence,that it had to control the cards used in its machines to avoid “injury to the reputation of the machines and the good will of” IBM. SeeInternational Business Machines Corp. v. United States, 298 U.S., at 139 -140.

8 Kodak, 504 U.S. at 485

9 The Court accepted the facts ISOs presented to prove that: (1) Kodak adopted its parts policy only after an ISO won a contract withthe State of California; (2) Kodak allowed its own customers to service their machines; (3) Kodak customers could distinguishbreakdowns due to poor service from breakdowns due to parts; and (4) many customers preferred ISO service.

10 Kodak, in fact, argued that it would have been ruinous for the company to increase prices for the monopolised provision of parts andservices since this would have lowered the number of machines purchased.

11 Court of Appeals for the Ninth Circuit no. 90-1029, page 11424.


68

The inventory cost explanation was judged inconsistent, since the main driving factor,the breakdown rates, seems to be unrelated to who is servicing the machines. In responseto this argument, moreover, the Court argued that this argument failed to explain whyKodak had to force parts manufacturers, equipment owners, and parts brokers not to sellto the ISOs since these actions would have no effect on Kodak’s inventory costs.

The free riding argument was also rejected. Kodak did not dispute that respondentsinvested substantially in the service market, with training of repair workers and investmentin parts inventory. Instead, according to Kodak, the ISO’s were free-riding because theyhad failed to enter the equipment and parts markets. Agreeing with the Court of Appeals,the Supreme Court acknowledged the perniciousness of this argument and concludedthat “one of the evils proscribed by the antitrust laws is the creation of entry barriers topotential competitors by requiring them to enter two markets simultaneously.”12

2 Discussion

This case presents us with two relevant economic issues associated with bundling andwe discuss each in order. First, we discuss the use of tying as a metering device.Second, we discuss the use of tying to create barriers to entry. We believe that both ofthese effects could have motivated Kodak’s behaviour.

METERING

One motivation for Kodak’s parts policy could be to act as a metering device. In general(as introduced in Section 3.3.1 and fully explained in Sections 5.5 and 5.6 of the mainreport), a firm is able to increase its profits by charging high value consumers more thanlow value consumers.13 In copiers, one way to distinguish between customers withdifferent values is to charge a usage fee for the copier. Possible methods for this usagefee include renting the copier and charging a price for every copy made, tying paper tothe machine and charging a price for the paper that includes a per-page copy fee, tyingtoner to the machine, and tying service to the machine. In this later scheme, to theextent the rate that a machine requires service is a direct function of the amount of usethat machine gets, Kodak is able to employ a metering fee on its copiers. That is, tyingservice to the equipment is a way of enabling Kodak to price discriminate amongcustomers according to the intensity of their use of copiers.

In general, consumers resist efforts of firms to price discriminate. For example, if Kodakhad tried to meter the usage of its copiers by tying its own brand of copier paper to thecopier, it probably would not have been successful. Since paper is interchangeable, itwould be difficult to enforce the tie without constantly monitoring the type of paper thatis used in the copier.14 This consumer behaviour points out an inherent inefficiency

12 504 U.S. 451, Section IIIB.

13 The alternative pricing structure is to have a single price, and the firm is left with the choice of charging a high price and only serving thehigh value customers, or charging a low price and serving all of the customers, but getting less revenue from the high value customers.

14 Enforcement issues could be weakened if machines were only leased and not sold. Rental only policy would be a way to prevent theexistence of a secondary market for used machines and parts. It would eliminate the availability of parts from the original manufacturersand ensure that the equipment producer would be the only company able to provide servicing to its machines. Such a rental onlypolicy, however, may meet with market resistance. Since Kodak faced competition for the sale of copiers, if it tried to impose a rentalonly policy, the high-value consumers would be likely to purchase from competitor copier companies, thereby defeating Kodak’s attemptto price discriminate. Moreover, rental only policies might well fall afoul of competition authorities in the presence of market power.


69

associated with price discrimination. High-value consumers will expend effort to appearto be low-value consumers so they can pay a lower price for the product. Thus theconclusion that price discrimination tends to be welfare enhancing when it expandsoutput must be tempered by the observation that price discrimination may have negativewelfare implications associated with high-value consumers trying to appear as low-valueconsumers (this point is expanded in Section 5.7 of the main report).

Given that Kodak was trying to meter its customers’ use, we can consider whether therewere alternative methods for Kodak to achieve a similar outcome. For example, insteadof refusing to sell parts to ISOs, Kodak could just have sold these parts at very highprices. Charging a high price would have also addressed Kodak’s concern that ISOs werefree riding on its investments. By charging high prices for its patented parts, ISOs wouldpay for all of the costs incurred by Kodak in developing its equipment. Kodak could havekept the overall cost of its copiers competitive by offering inexpensive maintenanceservice contracts to its customers and simultaneously charging high prices for parts.Kodak could have chosen its maintenance price mix to provide services competitivelywith respect to ISOs. This would have allowed Kodak to engage in a margin-squeeze onthe ISOs. The ISOs would be at a cost disadvantage to Kodak and Kodak could not havebeen accused of refusing to sell parts to the ISOs.

This strategy might have been circumvented if ISOs developed a way to fix broken partsin the equipment.15 Furthermore, to the extent the ability to charge high prices on partsand service is restricted by the competition for the original equipment, if Kodak were tocharge high prices for the provision of parts, the prices of the original copier would fall.As the price of new copiers fall, ISOs could have bought new machines to dismantlethem and use the constituent parts for servicing customers’ machines.

Refusal to sell parts to ISOs also can potentially be circumvented. For example, the ISOscould ask their clients to purchase the parts from Kodak and then resell these parts tothe ISO. This would be a more costly procedure than the ISO purchasing directly fromKodak. In fact, ISOs did buy parts from customers of Kodak that were willing to buyparts from Kodak and resell them to the ISOs.16 During the trial, first Kodak argued that itdid not require owners to buy service in order to receive parts.17 Kodak did admit,however, that it would sell parts only to those owners that agreed not to allow the ISOsto service the machines. The ISOs produced two statements made by third parties thatKodak suppliers had refused to sell them parts because of “arrangements with Kodak”or because the parts were “proprietary to Kodak.”18 Indeed, over time, Kodakincreasingly policed its no-parts sales policy. The company began tracking parts sales toensure that equipment owners were not purchasing more than they would reasonablyneed. It also required parts purchasers to provide proof of ownership of the preciseequipment model for which they were ordering parts. Furthermore, for copier parts, it

15 In other words, if when an ISO replaces a broken part in a Kodak copier it can refurbish this part and resell it to a later customer, thenthe ISO is able to limit the demand for Kodak replacement parts.

16 See Kodak II, 504 U.S. at 458 n.2

17 Kodak I, 903 F.2d at 615.

18 Kodak, 1988 WL 156332, at *2


70

demanded certification that the customer had a Kodak-trained employee to effect therepairs. The company was perfectly aware of the effect of their choice not to sell partsto ISOs, as an internal document states: “Both IBM and Xerox will sell spare parts, butwe do not. This makes it more difficult for a third party to service our copiers.”19

Kodak tried to persuade the Court that tying servicing to equipments did not make anydifference from the point of view of consumer welfare. The company claimed that allprofits coming from the provision of service on its copiers would be competed away atthe time of the original purchase of equipment.

This argument is correct if consumers are endowed with perfect foresight when thecontract is signed. If it is possible to perfectly anticipate what the cost and benefitsassociate with the machinery bought will be, such as the incidence and rate of breakdownsand the rate of depletion of the machinery, then tying is not an issue. Such an evaluationof the life-cycle costs of durable equipment is difficult for consumers to carry out.20

In addition, life-cycle cost evaluation needs to take into account the chance of Kodakchanging its policy on the provision and pricing of its parts and service. Only if customerscorrectly evaluate the cost of future parts and service of the copier will competition inthe market for copying machine offset the potential losses associated with tying.21

Even if consumers are able to evaluate the life-cycle cost of a new copier, there still canbe one-time losses associated with Kodak’s policy. Kodak refrained from explicitlyimposing tied service to customers who had already purchased their copiers prior to1985. While it appears that this strategy would have no effect on the existingconsumers, these consumers could still be harmed if the policy leads to the eliminationof alternative service providers. Over time, the number of older copiers in servicedeclines and the amount of that business is not sufficient to keep the ISOs in business.22

Soon, there will not be sufficient ISOs to service the machines that were sold before1985 and 1986 even though parts for these copiers were available to the ISOs. Hence,Kodak may be able to extract profits from these customers of the older copiers as aresult of its change in parts policy.

ENTRY DETERRENCE

The new Kodak policy could also have long-term negative consequences on the copiermarket. Kodak’s policy would effectively drive ISOs out of the market for providingservice to Kodak copiers (this motivation for bundling is discussed in Section 4.4.2 of the

19 Image Technical Service, Inc., et al. v. Eastman Kodak Co., C 87-1686 (January 18, 1996), Trial Exhibit 649 at 1315-1316, a Kodakinternal document. Similarly “The difficulty of obtaining parts, technical information and diagnostic software has effectively kept thirdparty service suppliers out of the advanced equipment service market,” Trial Exhibit 99 at 6762 (ibid.), a Kodak report on micrographicservice.

20 A way to make the life-cycle costing easy for consumers is for Kodak to rent its machines and charge a per-page fee for usage.Consumers, though, have not embraced such a strategy, in part because it leads to high-value consumers paying more for the copierservices.

21 If prices are public and not individually negotiated, then all customers would not need the ability to evaluate these costs, just“enough” consumers. That is, as long as the marginal consumers can evaluate costs of all expected future repairs and maintenance,then all consumers are protected from exploitation.

22 Whether this would lead to consumer harm is driven by the speed at which the competitive service providers are driven out of themarket relative to the length of the life of these older copiers.


71

main report). If the other copier companies followed suit, this would lead to theelimination of the entire ISO industry - only the copier companies would service theirown brand of copier. This would affect the ability of a new firm to enter the copiermarket. An entrant to the copier market would have to not only incur all of the costs ofdesigning, manufacturing, and selling copiers, but it would also have to set up an entirenetwork of repair shops to provide service for these new copiers. Clearly this createseven higher barriers to entry than just entering the copier market (and obtainingmaintenance services from ISOs). As we mentioned earlier, the Supreme Courtrecognised this effect when it concluded “one of the evils proscribed by the antitrustlaws is the creation of entry barriers to potential competitors by requiring them to entertwo markets simultaneously.”23

Not only can the raising of entry barriers lead to the potential ability for a firm tounilaterally raise price, these barriers to entry also can help facilitate a collusive outcome.Without the threat of new entry, the existing competitors in the copier market might beable to reach a coordinated outcome, either through tacit collusion or by explicitlyagreeing to a higher price. Thus, by raising entry barriers, Kodak not only would allowitself to act in an anticompetitive manner, but it also could lead to the entire industrybecoming more anticompetitive.

3 Conclusion

We believe that Kodak’s behaviour was motivated by its desire to better price discriminateamongst its end-users as well as the possibility that the behaviour would lead to a lesscompetitive copier industry over the long-run. To believe that the tying would beirrelevant over the short-run, one must trust that consumers possess sufficient informationto evaluate the life-cycle cost of new equipment. We are somewhat sceptical thatconsumers have this ability. Moreover, the dynamic consequences of the elimination of acompetitive servicing market are a cause for concern. Hence we believe that Kodak’schange in policy in terms of making parts available to ISOs was likely harmful to consumersin the short run and was also likely to be harmful to competition in the long run.

23 504 U.S. 451, Section IIIB.


72

VIIAspen caseIn this chapter we discuss a particular type of bundling - the bundling of substituteproducts as a means to decrease the demand for a competing product. The case arisesfrom the complaint made by the skiing facilities operator Aspen Highlands Skiing Corp.(Highlands) in 1979. Highlands, which owns one of the four major mountain facilities for downhill skiing at Aspen, Colorado, filed a suit in the Federal District Court ofColorado against Aspen Skiing Co. (Aspen Ski), which owns the other three major skifacilities in Aspen, alleging that Aspen Ski had monopolised the market for downhillskiing services in Aspen, in violation of Section 2 of the Sherman Act. The basis for thecourt decision in this case was that Aspen Ski was an essential facility - a finding thathas caused much consternation amongst economists in the ensuing years. This case hasled to much controversy in later cases, such as the Microsoft cases in the U.S., thathave made allegations about what is and is not an essential facility. We show that interms of economic analysis, one does not have to rely upon the concept of essentialfacilities to show that the behaviour by Aspen Ski of bundling substitute products leadsto anti-competitive effects that harmed competition.

1 Background1

Between 1945 and 1960, three private investors independently developed three majorfacilities for downhill skiing at Aspen: Aspen Mountain (Ajax), Aspen Highlands(Highlands), and Buttermilk. A fourth mountain, Snowmass, opened in 1967. From 1958,each of these independent companies offered lift tickets for the use of its ownmountain. In 1962, a 6-day, all-Aspen ticket was introduced. The 6-day pass allowedskiers who visited the resort for weekly periods to be flexible about where to go skiingeach day. The all-Aspen ticket program consisted of booklets containing six coupons,each redeemable for a daily lift ticket at Ajax, Highlands, or Buttermilk. The price of thebundle was discounted compared to six single-day tickets, and all of the coupons had tobe used within a week. The revenues from the sale of this ticket were distributed inaccordance with the number of coupons collected at each mountain.

In 1964, Aspen Ski purchased Buttermilk and offered a 2-area, 6-day or 7-day tickets incompetition with the 3-area booklet. Three years later, Aspen Ski opened the fourthmountain, Snowmass. Aspen Ski offered a 6-day ticket bundle that was good only at thethree Aspen Ski mountains. This offer by Aspen Ski was outsold by the all-Aspenmountains pass.

1 Most of the information for this case study is taken from the decision of the U.S. Supreme Court in Aspen Skiing Co. v AspenHighlands Skiing Corp., 472 U.S. 585 (1985). This decision affirms the decision in Aspen Skiing Co. v Aspen Highlands Skiing Corp. ofthe Court Of Appeals For The Tenth Circuit. No. 84-510. The District Court decision is 738 F.2d 1509, District Court of Colorado.


73

In the 1971-1972 season the all-Aspen ticket changed and skiers no longer had to visitthe ticket window every day to gain access to the slopes. The ticket consisted of a cardworn around the skier’s neck that was checked at the chair lifts. In that season,operators at Highlands monitored usage of the ticket by recording the ticket numbers ofpersons going onto the slopes of that mountain. There was some concern that usage of the all-Aspen ticket should be monitored by a more scientific method. After a one-season absence, the pass was reintroduced with a new method of allocating revenuesbased on usage. Lift operators punched the ticket when the skier first sought access tothe mountain each day. Random-sample surveys were commissioned to determine howmany skiers with the 4-area ticket used each mountain, and the parties allocatedrevenues from the ticket sales in accordance with the survey’s results.

During the following 4 seasons Highlands’ share of the revenues from the ticket rangedfrom 18.5% to 13.2%. At that time, Aspen Ski was not offering its own 3-area, multi-dayticket in competition with the all-Aspen pass. The all-Aspen pass represented 35% of thetotal revenue for the ski areas.

Until 1977, Aspen Ski and Highlands had independently offered passes of various length(from one day to six days) at each of their own mountains. For the 1977-1978 season,Aspen Ski imposed as a condition to continue the all-Aspen ticket that Highlandsaccepted a 13.2% fixed share of the tickets revenues. The figure corresponds to thelowest Highlands’ share of revenues in those years, and was the result it achieved in aseason of scarce snow and an unusually low number of visitors. Highlands wanted tocontinue to divide revenues on the basis of actual usage as estimated by the annualsurvey of skiers. After negotiations Highlands accepted a fixed percentage of 15% forthat season. Aspen Ski reinstated its 3-area, 6-day ticket during the season, but thatticket was outsold by the all-area pass by nearly two to one.

The Supreme Court discusses the evidence on Aspen Ski management’s views on thepass during the 70’s. The company complained that the coupon method of monitoringusage was administratively cumbersome. Also, Aspen Ski doubted the accuracy of thesurvey and complained about the appearance of the college students who wereconducting the survey. Aspen Ski’s president also expressed the view that the existenceof the 4-area ticket was draining off revenues from Aspen Ski.2 Consequently, in 1978,Aspen Ski management recommended to the board of directors that the 4-area ticket bediscontinued. The board decided to offer Highlands a 4-area ticket provided thatHighlands would agree to receive a 12.5% fixed percentage of the revenue.3

Highlands suggested alternative ways to distribute revenues based on usage to bemonitored by coupons, electronic counting, or random sample surveys. Moreover, toaddress Aspen Ski’s concern about how the survey was run, Highlands proposed to hiredisinterested ticket counters at its own expense to count or survey usage of the 4-areaticket at Highlands. Aspen Ski refused each of these methods of measuring usage andeventually Highlands rejected the offer of a fixed percentage.

2 472 U.S. at 586-587, 950, 960.

3 Later in the season a member of Aspen Ski’s board of directors informed a Highlands official that he had advocated making Highlandsan offer that [it] could not accept. Id., at 361.


74

At the same time, Aspen Ski actively engaged in an advertising campaign that stronglysuggested to the public unfamiliar with Aspen that its mountains were the only ones inthe area.4 Aspen Ski also hindered Highlands’ effort to market its own multi-areapackage to replace the joint offering. The company refused to sell Highlands any lifttickets, either at the tour operators’ discount or at retail. When Highlands finallydeveloped an alternative all-area pass, called the “Adventure Pack,” which consisted of a 3-day pass at Highlands and three vouchers5, each equal to the price of a daily liftticket at a Ski Co, Aspen Ski refused to accept them. Later, Highlands redesigned theAdventure Pack in such a way that Aspen Ski eventually accepted it, but the AdventurePack met considerable resistance from tour operators and consumers accustomed to the convenience and flexibility provided by the all-Aspen ticket.

Without a convenient all-Aspen ticket Highlands became a one-day ski area. Its share ofthe market for downhill skiing services in Aspen declined steadily after the 4-area ticketwas abolished: from 20.5% in 1976-1977 to 11% in 1980-1981. The associated skiingservices revenues (like ski school, ski rentals, amateur racing events, and restaurantfacilities) declined sharply as well.

Aspen Ski discontinued the 3-day, 3-area pass for the 1978-1979 season, with negativeconsequences on the demand for Highlands. With the three-day ticket, a person couldski on the Aspen Ski mountains for three days and then have three days in which to skion Highlands.6 Due to strong consumer demand, Aspen Ski reinstated the 3-day ticketlate in the season, but without publicity or a discount off the daily rate.7

2 Court decisions

The Tenth Circuit Court of appeals affirmed the decision of the District Court of Coloradobased on liability under the essential facility doctrine. The Court found sufficient evidenceto conclude that the defendant acted with the intent to create or maintain a monopoly. In its review of the evidence on the question of intent, the Court of Appeals noted thatby refusing to cooperate with Highlands, Aspen Ski “became the only business in Aspenthat could offer a multi-day multi-mountain skiing experience”8; that the refusal to offer a 4-mountain ticket resulted in “skiers’ frustration over its unavailability”9; that there was no valid business reason for refusing to accept the coupons in Highlands’ AdventurePack; and that after Highlands had modified its Adventure Pack to meet Aspen Ski’sobjections, Aspen Ski had increased its single ticket price to $22 “thereby making itunprofitable... to market [the] Adventure Pack.”10

4 For example, Aspen Ski had a sign changed in the Aspen Airways waiting room at Stapleton Airport in Denver. The old sign had apicture of the four mountains in Aspen touting “Four Big Mountains” whereas the new sign retained the picture but referred only tothree mountains.

5 The vouchers were guaranteed by funds on deposit in an Aspen bank, and were redeemed by Aspen merchants at full value.

6 Id., at 245.

7 Id., at 622.

8 Joint appendix on appeal prepared by the parties, I App. 191, 243.

9 Quote from joint appendix on appeal prepared by the parties, II App. 525-33.

10 Aspen Ski set its single ticket price at $22 and discounted the 3-area, 6-day ticket to $114. Quote from the original record, IV R. 534-35; IX R. 1424-25.


75

In the Supreme Court, Aspen Ski contended that even a firm with monopoly power hasno duty to engage in joint marketing with a competitor. It also contended that the Courtof Appeals incorrectly relied on the essential facilities doctrine. In response, Highlandssubmitted that, given the evidence in the record, it was not necessary to rely on theessential facilities doctrine in order to affirm the judgment. The Supreme Court affirmedthe decision of the Court of appeals based solely on the intent test. The Courtspecifically found it unnecessary to consider the possible relevance of the essentialfacilities doctrine.

The Supreme Court concluded that consumers were adversely affected by theelimination of the all-Aspen tickets. Surveys, expert testimony and other evidencesupported this conclusion.11 The actual record of competition between a 3-area ticketand the all-Aspen ticket in the years after 1967 indicated that skiers demonstrablypreferred four mountains to three.12 Highlands’ marketing expert testified that many ofthe skiers who come to Aspen wanted to ski the four mountains, and the abolition of the4-area pass made it more difficult to satisfy that ambition.13 The Court ruled that therewas evidence of an adverse impact on Highlands from Aspen Ski’s strategy. Thisevidence was based on Highland’s unsuccessful attempt to create similar packages tothe all-Aspen package and the steady decline of Highland’s share of the relevant marketafter the ticket was terminated. Finally, it was believed that Aspen Ski failed to providean adequate business justification for its behaviour. Instead, the court concluded, AspenSki sought, even forgoing revenue, to reduce competition in the market over the longrun by harming its smaller competitor.

3 Economic analysis

Seen from a traditional perspective, the conclusion that Aspen Ski, by excluding Highlands,was seeking to reduce competition would appear to be a logical contradiction. AllowingHighlands into the multi-mountain pass is a form of cooperation or even collusion. OnceHighlands is part of the pass, the two companies have agreed not compete on price (at least for this significant part of the market). Thus if excluding Highlands is denying itan opportunity to collude, it is hard to see how this would be reducing competition. One significant source of competition is that even with collusive pricing, the two firmsseek to gain a greater share of the skiers and thus a greater share of the joint ticketrevenue. This competition for share of the bundle, while it might not be pricecompetition, could still lead to significant benefits to consumers.

Dennis Carlton (2001) makes the argument that the court case was simply an attempt byAspen Highlands to get a better deal from the courts than it could achieve on its own inthe negotiation with Aspen.14 We think he is correct in this assessment, although we do

11 Record Ex. No. 75, pp. 36-37.

12 472 U.S. 589-590, 592. On this point see also the view of a major wholesale tour operator who asserted that he would not evenconsider marketing a 3-area ticket if a 4-area ticket were available. 472 U.S. 585, 607.

13 Tr. 775.

14 See Dennis Carlton, 2001, “A General Analysis of Exclusionary Conduct and Refusal to Deal-Why Aspen and Kodak are Misguided,”NBER working paper 8105.


76

not agree with the conclusion that he then draws. Carleton concludes that since a four-mountain pass is the efficient outcome, we can expect the parties to reach thisagreement and the courts should not get involved.15 The fact that Aspen Highland’sbargaining power may not be proportional to its fraction of skiers is not a problem thatthe court should try to rectify.

Our view is that Aspen took actions akin to predatory pricing. Its bundled pricing of multi-day discounts, the temporary elimination of the 3-day ticket, and exclusion of Highlandswas a way to reduce consumers’ willingness to pay for Highlands to a level well belowwhat they would have paid in a market with no bundled sales or discounts. This actionwas costly to Aspen Ski and detrimental to its customers. As we explain in more detailbelow, the exclusion was most likely an attempt by Aspen Ski to improve its bargainingposition - to put Highlands into a position of financial distress so that it could bepurchased on the cheap by Aspen Ski.16

It is possible that if Aspen Ski were to merge with Aspen Highlands, the four-mountaingroup would raise prices. Existing homeowners and hotels in Aspen might becomesubject to a monopoly. But we do not think that is the primary problem. Aspen mustcompete with other resorts such as Vail and Park City and this limits even a four-mountain group’s ability to raise prices. Furthermore, if this were the concern, thenallowing Aspen Highlands into the group pass would lead to just this monopoly outcome.The merger, though, would lead to the loss of competition between Highlands andAspen Ski for share of the bundle.

The theoretical analysis of this case, that of ex-ante complements and ex-postsubstitutes, is presented in detail in section 4.4.4 of the main report. Here weemphasise how Aspen Ski used the variety bundle to create an artificial advantage.Consumer demand for skiing at each of the individual mountains is likely to be positivelycorrelated (if a consumer has a high value of skiing at Ajax, that consumer is likely to alsohave a high value of skiing at Highlands). Consumers also value variety. This is especiallytrue for customers who are considering the Aspen area over other resort options. Thesetwo assumptions on preferences will yield the types of demands described in the case -the all-mountain pass is significantly more popular than the pass offering only threemountains, both for skiers and for travel agents.

For simplicity, we can assume that the 4 mountains are of similar quality, so if only single-mountain passes are sold, each mountain should get approximately 1⁄4 of the business.While the 4-mountain bundle was the most-preferred item, a 3-mountain pass would bemore expected to capture more than 3⁄4 of the market against a single mountain of equalquality. This is because in looking at the value a consumer places on a bundle containing theoption to ski at multiple mountains, there is diminishing valuation to variety. The consumer

15 Carlton makes an exception for cases where the bundle is not provided by a single firm. In this case, since Aspen Ski was the result ofa merger, one could extend Carlton’s exception to include restricting this type of behaviour as part of the merger condition.

16 We note that other potential buyers would expect to face the same problem competing against the Aspen Ski bundle and thus wouldnot offer competition in the purchase of Highlands. The courts might be unlikely to prevent such a purchase both due to the weakenedfinancial state of Highlands and the recognition that Aspen is competing in a larger market against other resorts such as Vail or ParkCity.


77

values the second mountain added to the pass at less than the first, and the third at lessthan the second, and so on. The Aspen 3-mountain pass turned Highlands into themarginal fourth mountain rather than one of four mountains creating a bigger pie.17

Highlands did everything it could to create its own version of a 4-mountain pass. As weknow from the previous sections, Highlands was ready to buy Aspen Ski tickets toinclude them in the new all-areas “Adventure Pack.” Aspen Ski was willing to forgo dailyticket sales to skiers who bought the Adventure Pack. Such behaviour makes sense onlyif Aspen Ski believed that many of these skiers would buy tickets at Aspen Skimountains if the Adventure Pack did not include the Aspen Ski passes. By restrictingHighlands to only selling passes to its own mountain, Aspen Ski was able to limit thesize of the market available to Highlands. Since consumers value additional mountain atdecreasing amounts, Aspen Ski can price its 3-mountain pass at such a level thatHighlands will not be able to get many customers. Since the costs associated withrunning a ski area are primarily fixed, losing customers would eventually drive Highlandsout of business or, more likely, force Highlands to sell itself to Aspen Ski at a lower price.

For the 1981-1982 season, Aspen Ski set its single ticket price at $22 and discounted the3-area, 6-day ticket to $114 or $19/day. Using these numbers and assuming that theindividual mountains on average are valued equally by consumers, we can see the bindHighlands is facing. The increment Highlands can charge for the 4-mountain pass abovethe 3-mountain pass is small.18 If there is no increment, then Highlands charges $114 forthe 4-mountain, 6-day pass. Assuming consumers value each of the mountains equally,then on average each consumer will ski 4 or 5 days at Aspen Ski and 1 or 2 days atHighlands. Highlands is paying Aspen Ski $22 for each day the consumer skies at AspenSki, so for the day (or two) that the consumer skies at Highlands, the revenue earnedwill be $4 if only one day is skied there, or $13 per day for each day if two days are skiedthere. In either case, the revenue earned by Highlands is well below the $22 per day thatAspen Ski is earning.

We conclude from the evidence that Aspen Ski was refusing to deal with Highlands inorder to gain a competitive advantage over Highlands.19 Such a refusal to deal has harmto consumers if it leads to Highlands going out of business (or selling itself to Aspen Ski).Therefore, there should be some remedy imposed to ensure that there is fair competition.The goal of the remedy is to ensure that Aspen Ski is not able to disadvantage its rivalthrough a bundled pricing strategy.

One such remedy would be to allow Aspen Ski to set its own price for the 3-mountain passand then require Aspen Ski to let Highlands participate in this pass, earning the averageprice per day for every day that the consumer chooses to ski at Highlands. Table 1 presentsthe prices for the 2001/2002 ski season at Aspen Ski when the tickets are bought in advance.Table 2 presents the same information for purchases that are made at the mountain on

17 One could argue that if the customer bought the Aspen Highlands ticket first, then the three Aspen mountains would become themarginal 2nd, third, and fourth mountains, but this would be a case of the tail wagging the dog. Few people come to Aspen withoutexpecting to ski Aspen, Buttermilk, or Snowmass.

18 We do not know exactly what this increment is, so for this example we will assume it is zero and the numbers presented are then aworse case scenario for Highlands.

19 Given the technology available today, there is no reason that there should be any extra costs with monitoring the number of days amulti-day pass holder skies at each mountain, so the concerns Aspen Ski voiced over the reliability of the survey is not relevant today.


78

the day that the skiing starts.20 Using these numbers as a guide, Aspen offers a $7/dayadvance-purchase discount which leads to a price of $57 per day to ski at its mountains.Thus, Highlands should be allowed to offer a 4-mountain pass at the price of $57 per day.There can be card readers at the chair lifts that monitor the usage of each mountain, andthe revenues can be split between Highlands and Aspen Ski according to the actual usageat each mountain. Such a solution would result in each of the mountains competing onits own merits. If a mountain wanted to offer a particular deal on a single-mountain pass,then such a pass can be offered completely independently of the multi-mountain pass.

TABLE 1

2001/2002 7-DAY ADVANCE PURCHASE PRICES

Length of Pass Adult

(Ages 18-69)

7-day $399($57/day)

6-Day $342($57/day)

5-day $285($57/day)

4-day $228($57/day)

Source: https://store.aspensnowmass.com/

TABLE 2

2001/2002 LIFT TICKET PRICES PURCHASED ON THE MOUNTAIN

Pass Length Adult Child Youth Senior

(Ages 18-69) (Ages 7-12) (Ages 13-17) (Ages 65-69)

7-day $448 $273 $343 $413($64/day) ($39/day) ($49/day) ($59/day)




Source: https://store.aspensnowmass.com/

20 According to these prices, we note that Aspen no longer offers discounts for multi-day passes. Now that the court has forced it toinclude Highlands in the multi-day pass, there may be less incentive to offer a multi-day bundled ticket discount.


79

VIIIGuinness and GrandMetropolitan mergerOn 12 May 1997, the Boards of Directors of Guinness plc (Guinness) and GrandMetropolitan plc (Grand Met) announced a planned merger of the two companies. The merged firm saw five primary benefits from the merger: (i) complementary productportfolios; (ii) greater geographic breadth; (iii) increased marketing ability; (iv) costefficiencies; and (v) increased financial resources for future expansions. The firmbelieved that the merger would provide the new company with significant economies of scale and scope. This transaction was reviewed by both the European Commissionand the U.S. Federal Trade Commission, and before being consummated there were anumber of conditions that had to be met. In this chapter, we present the factualbackground and analysis that underlie these remedies. We also present an alternativeanalysis of the facts that imply an alternative remedy that would have been less costly to the firms yet fully protected consumers and competition from adverse effect from the merger.

1 Background

Guinness

Guinness was originally a brewer of beer but had diversified itself into a producer ofspirits worldwide. Guinness began in 1759 when Arthur Guinness took over an unusedbrewery at the periphery of Dublin. By 1980, Guinness Stout, the flagship beer brand,was brewed in 50 countries and sold in 150 different countries. In 1985 and 1986Guinness acquired Distillers Company Limited and Arthur Bell & Sons. With thesepurchases, Guinness became a holding company for United Distillers - a producer anddistributor of spirits worldwide. The spirit brands controlled by Guinness includedJohnnie Walker, Gordon’s, Bell’s, and Dewar’s. In addition, Guinness owned 34% ofMoët Hennessy, a leading producer of champagne and cognac.

Grand Met

The Grand Met group of companies was involved in the production and distribution ofspirits, food manufacturing, and ownership of fast-food restaurants. The companyoriginated with a hotel business called MRMA, which became Grand Metropolitan HotelsLtd in 1962. It entered the catering industry by acquiring Express Dairy Ltd in 1969 andBerni Inns Ltd in 1970. Grand Met made further acquisitions in the entertainment and


80

drinks businesses including the purchase of Mecca (bookmaking business) and thebrewery companies Truman, Hanbury and Buxton Ltd, and Watney Mann Ltd. The 1980sbrought further growth, notably by the acquisition of wine and spirits assets of theLiggett Group Inc, Warner Holidays Ltd, a group of three hotels in Paris, InterContinentalHotels Corporation, Heublein Inc (wines and spirits), and several other United Statesretailing and service companies. In the second half of the 1980’s Grand Met focusedmore on its food, drinks, and retailing businesses, especially branded goods andservices, and disposed of a number of the businesses, including Intercontinental Hotels.Acquisitions in the selected areas of activity have continued with William Hill OrganisationLtd (in bookmaking) and The Pillsbury Company, a leading United States food processingcompany which owns the fast-food operation Burger King among other interests.

The merged company, taking the name Diageo, organized itself into four divisions: UDV,Pillsbury, Guinness and Burger King. The spirits and wines businesses are controlledunder the UDV division. The revenue of the UDV division in 1996 was 10 billion ECU,which was 47% of the revenue of Diageo.

2 European Commission decision

MARKET DEFINITION

The Commission concluded that every type of spirit constituted a separate productmarket. The Commission further determined that branding and brand name were keycompetitive characteristics in these product markets. The Commission also concludedthat the relevant geographic markets were national despite arguments from Guinnessand Grand Met that the correct geographic market would encompass all of the EEA.Justification for this definition included that the per capita product consumption variedbetween the Member States, the distribution of these products was organized on anational level, and there was not complete free trade between countries of theseproducts. Also, there are major differences in taxation and regulation of these productsbetween the different member states.1

The Commission also considered whether the country in which the spirit is made is animportant factor for defining relevant markets. Whiskies provide an example of thisissue. Scotland requires that in order for a malt to be marketed as a Scotch Whisky, thewhisky must be wholly distilled within Scotland and matured there for at least threeyears. The Commission decided that for the purposes of this case “the issue ofsegmentation of whisky by origin is only analysed further where the different possibledefinitions would lead to significantly different shares and overlaps (notably, Spain andIreland).”2

1 This does not hold for duty-free sales, since consumers are travelling across borders.

2 European Competition Commission Decision of 15 October 1997 “Declaring a concentration to be compatible with the commonmarket and the functioning of the EEA Agreement,” Case No IV/M.938 - Guinness/Grand Metropolitan (98/602/EC).¶16


81

VERTICAL ISSUES

The Commission found that the distribution systems within the markets demonstratedthe importance of vertical integration. Wholly owned subsidiaries were responsible forthe exclusive distribution of products at a national level. In addition, to distributing theirown brands, the distributing subsidiaries of Guinness and Grand Met would distributespirits of other firms. This third-party distribution accounted for only a relatively smallproportion of total distribution activity, so the distribution of other firms’ products wasviewed as only a source of incremental profits to the distribution of one’s own products.Since this third-party distribution was not considered to be a critical asset of these firms,there could be a concern that it might be dropped if it would harm the competition.“Integration into distribution allows the supplier to maintain control over the marketingand distribution of the brands which it owns, thereby safeguarding the all-importantimage of those brands in the market-place.”3

PORTFOLIO EFFECTS

The Commission found that portfolio effects were present since the holder of a portfolioof leading spirit brands may enjoy a number of advantages. In particular, the Commissionfelt that the position of a supplier of a broad range of spirits to its customers would bestronger than a supplier of a single spirit since it would account for a greater proportionof the customers’ business. A large portfolio of products would provide greater flexibilityto structure prices, promotions and discounts, as well as provide a greater potential fortying. Moreover a broader portfolio provides economies of scale and scope in sales andmarketing activities and strengthens the force of a threat of refusal to supply.

The Commission established that the strength of these advantages, and their potentialeffect on the competitive structure of the market, depends on a number of factors, including:

• whether the holder of the portfolio has one or more leading brands in a particular market;

• the market shares of the various brands, particularly in relation to the shares ofcompetitors;

• the relative importance of the individual markets in which the parties have significantshares and brands across the range of product markets in which the portfolio is held;and/or

• the number of markets in which the portfolio holder has a brand leader or leading brand.

The parties’ competitors confirmed the existence of a portfolio effect in the market.4

While it was true that end consumers purchase only a specific brand of spirit, not broadportfolios, the manufacturers often sell all of their products to intermediate retailers (or bar owners).5 It is this portfolio of products that is important to the intermediaries.One reason for this might be that if there are two different mid-tier gin brands (for example),

3 Ibid., ¶20.

4 The Commission asked ten firms the question “does possession of a leading brand in all or most spirit categories help sales of spiritsin general?” and eight firms responded that it would help a lot.

5 Generally there are two types of outlets that purchase spirits from manufacturers. Off-trade, which are liquor stores and other types of retailers, that sell packaged goods to be consumed away from the retail establishment, and on-trade, which are pubs, restaurantsand nightclubs, that sell spirits for consumption on the premises.


82

a spirit retailer perhaps is more likely to buy the one that is distributed by the firm thathas the leading brand of some other spirit (assuming the two gin brands are comparablypriced) as this purchase will minimise the number of distributors that call on that retailer.Thus, by expanding the number of leading brands it distributes, the merged companywould be in a position to get additional sales of its secondary brands. The Commissionfound that the strength of the portfolio effect varied from market to market. The oneplace in which it was a significant problem at the time of the merger was Greece.

The decision states that portfolio power can also be seen as a barrier to entry. There aretwo ways these barriers to entry arise. First, a new entrant would need to develop arange of desirable products in order to counter the portfolio power of the merged company.Without such a portfolio, a retailer might feel threatened to a potential cut-off of supplyby the firm with portfolio power if the new entrant’s products are purchased. Second,the vertical integration of the major industry players limits the supply of third-partydistributors to assist new entrants. If there are not many (or any) non-vertically integratedspirit firms, then there will be no distribution available to distribute the new entrant’sproducts. Even without these barriers, there are significant costs of entry as entrantswould need to promote their brands heavily, and sometimes pay a listing fee to encourageretailers and wholesalers to take the new product. Also, the importance of the on-tradedistribution channel makes new entry hard. On-trade consumption often is key in shapingconsumer demand for a product as customers are more likely to try a new product ifthey only have to buy a glass. The on-trade industry is more fragmented than the off-trade, though, and thus direct distribution to the on-trade would be extremely costly tothe new entrant. Thus, barriers to entry are generally very high in the spirits industry.

EFFECTS IN SELECTED MARKETS

Ultimately, we are interested in whether the remedy obtained by the Commission wasreasonable and whether there was an alternative remedy that would allow thecompanies to enjoy more of the efficiencies of the transaction without being able toharm competition. Therefore, we present the Commission’s analysis and remedy foreach of the affected countries. The Commission concluded that the merger had thepotential to create or strengthen a dominant position in the following countries:

Greece

Overall, the Commission had three broad concerns in Greece. The creation of a dominantposition in whisky, the creation of significant portfolio power, and the creation of barriersto entry in spirits through the use of the firms’ distribution facilities. Greece’s nationaldrink, ouzo, accounted for a large portion of national spirit sales. Whisky consumptionwas growing, and it was the most consumed imported spirit in Greece. The mergerwould have brought together the high market shares in rum and gin of Guinness withthe high market share in brandy of Grand Met. The merged company would have beenover four times as large as the next largest competitor in Greece, becoming the largestimporter and distributor. It would have supplied the leading brands in whisky, gin, rum,and brandy.


83

Spain

The Commission found that the merger would lead to the creation of a dominantposition in the Scotch whisky market. The merged company would have been over twiceas big as its nearest competitor (Allied Domecq) in Spain. The Commission pointed tothe scarcity of successful new entrants into the whisky market in the ten yearspreceding their investigation as evidence of significant entry barriers.

Ireland

The merger would combine the distributor Gilbeys of Ireland with 50% ownership ofCantrell & Cochrane. Thus the merged firm would have interest in undertakings with astrong combined share of distribution of vodka, rum, Scotch whisky, and brandy/cognac.The Commission was concerned that the merger would reduce the number of significantindependent distributors in Ireland to two, as well as strengthen the dominant positionsof the parties in Scotch whisky and brandy/cognac.

Belgium/Luxembourg

The Commission found that the merger would create market dominance inBelgium/Luxembourg in gin, vodka, and whisky.

PARTIES’ UNDERTAKINGS AND COMMISSION’S CONCLUSION

The parties agreed to the following undertakings to remedy potential areas of concernidentified by the Commission.

• Divest Dewar’s and Ainslie’s brands of Scotch whisky.

• Terminate the Wyborowa agency (vodka sales) in Belgium/Luxembourg.

• Appoint an independent third-party distributor for gin in Belgium/Luxembourg for atleast nine years.

• Discontinue the Bacardi rum agency agreement currently held by Guinness in Greece.

• Dispose of certain interests in Ireland in order to ensure competition in distribution.6

The Commission concluded that the effect of all of these divestments as well as otherundertakings were sufficient to address its concerns and allowed the merger to proceed.

3 Analysis

Many of the Commission’s concerns with this transaction arise from the combination ofhorizontal competitors. We will not examine these horizontal concerns but instead willfocus on the concerns raised by an alleged increase in portfolio power and an increase inbarriers to entry through the control of distribution assets. A definition of portfolio poweris given in the Commission decision: “in short the market power deriving from a portfolioof brands exceeds the sum of its parts.”7 This definition is in stark contrast to the

6 The identity of these divestitures is not included in the public version of the decision because of confidentiality concerns.

7 Ibid., ¶38.


84

Chicago School concept of there being a single monopoly rent - in such a world, there isno ability for there to be more market power than the simple summation of the marketpower in each individual market. There are four possible sources of portfolio power:

(i) From the retailer’s standpoint, if a firm has a sufficiently wide range of products, thenthe retailer might decide that it can save transactions costs by dealing with just asingle firm. This type of benefit is the same as the efficiency reason for bundlingdiscussed in Section 4.3.1 of the main report.

(ii) Second, a portfolio of products would provide the firm with economies of scale andscope. There are fixed costs of distribution and storage, and if these fixed costs canbe spread over more products then there are overall cost savings to the firm.

(iii) Third, a broad portfolio could increase the potential for tying and bundling. A wide rangeof brands provides opportunities to offer discounts on bundles of different products.Also, in theory, a firm could tie the availability of a leading brand in one market to thepurchase of a secondary brand in an alternative market. Another issue might be theability to have bundled pricing. This is discussed in Section 4.4.1 of the main report.

(iv) Finally, a possible threat to refuse supply would have its effect magnified by therange of products that are potentially going to be withdrawn.

We note here that the first two of these sources tend to be pro-competitive and the latertwo are likely to be illegal given the current competition laws around the world.

In the proposed merger, as it was originally structured, there was the opportunity forsignificant economies in the distribution of the spirit brands of the merged firm. Theseeconomies will lower the costs of bringing to the market all of the products of the firm.While this represents an advantage the merged firm’s products have over its rival’sproducts, we believe that the competition authorities should welcome such cost savingsas being beneficial to consumers. We do acknowledge that there are potential sideeffects from these advantages that need to be evaluated before allowing the merger togo forward. These side effects can be evaluated for their impact on two different typesof competitors: existing competitors and potential competitors.

For the existing competitors, if the cost advantages are so great that competitors will bedriven out of the market and once driven out unlikely to re-enter the market, then therecould be concerns of higher prices in the future. In the spirit industry we are not awareof any concern about competitors being driven out of the market, so we cannot evaluate thisconcern in the proposed merger. If this was a concern, then there needs to be a balancingof the benefits of the greater competition from the future loss. In particular, one shouldcompare the benefits to consumers arising from the competition that leads to the exit ofcompetitors with the harm to consumers from the future higher prices. This comparisonof benefits and harms also needs to be discounted over time.8 This type of evaluation isfurther explained in the discussion of the GE/Honeywell merger in Chapter IV above.

8 Such an evaluation might not have to be conducted if there is a fix that can be imposed on the merger that leads to the enjoyment ofthe benefits of greater competition without the future costs of higher prices.


85

A second potential concern from the distribution economies is that it becomes harder fora new entrant into the spirits industry to obtain distribution of its products. The economiesof scale in distribution make it unlikely that a new entrant can start with an efficientdistribution network on its own and if there in not an independent competitivedistribution network, then the new entrant might not be able to achieve the level of salesneeded to justify the risky investment of the new product. The mechanism here is thesame as the discussion of bundling to create an entry barrier in Section 4.4.2 of the mainreport. While we do not know if this was a concern in the spirit industry in the selectedcountries, it does appear to be a valid concern. One solution for this concern, though, isnot to force the merging firms to divest brands and hence lose the potential benefits ofthe combined distribution systems, but to require that the distribution of the merged firmnot be exclusive. With such a requirement, any new entrant can obtain the mostefficient distribution for its product (if the new entrant desires such distribution), andthus there are no anti-competitive barriers to entry that are erected in the spirits market.

In addition to the economies of scale in the distribution of spirits, there could well beother advantages arising to the merged firm from having a broader portfolio of spirits.One would have to investigate whether or not entry with a single brand of spirit wouldbe more difficult if the merger happened than otherwise. If so, then there would have tobe an effort to evaluate how the merged firm could still gain any benefits from themerger, but also be able to grant these benefits to competing firms.

Another issue in this proposed merger was the ability of the merged firm to have bundledprices. Such prices might put competitors at a disadvantage. For example, if the mergedfirm offered a discount that increased with the number of different brands that aparticular retailer purchased, then the merged firm’s products could have an advantageover its rivals. This type of bundling and its theoretical effects are discussed fully inSection 4.4.1 of the main report. As is shown in that Section, the firm must have marketpower in each of the bundled products and not be able to price discriminate amongst itscustomers for the bundled pricing to disadvantage rivals. Furthermore, such anadvantage is one of lower prices to consumers which in general should be encouraged.Observing that the merged firm would potentially have an advantage is not sufficient tojustify forcing the firm to sell off some of its brand. As in the distribution case, thereneeds to be an analysis that the advantage is likely to lead to exit of existing firms (or theinability of new firms to enter), and if firms exit they are unable to re-enter in the future.

In this case, the types of evidence we would like to have is information on the strengthof the brand preferences in particular markets, so we can evaluate whether a bundleddiscount is likely to be sufficient to overcome any existing preference for thecompetitor’s products. One source of this information would be the past response tobundled pricing of the various firms. If this bundled pricing did not lead to any anti-competitive effects, then we need to have a reason why the increased ability to bundlewould begin to have anti-competitive effects.9 Another piece of evidence we would like

9 In its decision, the Commission noted that there were a number of examples of bundled discounts in the past. What is not available inthe public version of the decision is what effect these bundled offerings had on the shares of the various competitors in the market.


86

to have is whether there is a single price to all customers or if there are negotiateddeals. When each customer obtains a different bundled price, then there are not thecompetitive advantages to bundling as in the case when there is a single market price.

If the conclusion of this analysis was that the bundled prices were likely to lead tocompetitive harm, then we do not believe that the appropriate remedy is to require thedivestiture of brands. A simple pricing rule that requires the firm to offer the same pricesregardless of the number of brands the retailer buys would eliminate any anti-competitive effect from the bundled prices. Normally the Commission has been reluctantto accept behavioural undertakings that purport to alleviate competitive concerns. This reluctance arises because of the ongoing costs of having to monitor the proposedbehaviour as well as a general concern that regulating a dominant firm is usually lessefficient from society’s standpoint than not having a dominant firm in the first place. In this type of bundling case, though, we believe that these types of concerns as to theremedy are not justified. First, the ongoing monitoring costs should be negligible as the customers of the firm are the natural enforcers. The rule (no bundled offeringsunless the constituent parts of the bundle are available to be purchased separately onthe same terms) is easy to evaluate and it is in the customer’s own best interest toenforce the rule. Therefore, the customers would be the de facto enforcement agency,not the Commission. Second, the remedy is being imposed because there are benefitsto the proposed integration that can only be enjoyed because of the merger. Society isoverall better off by allowing the merger and this benefit is not outweighed by the smallenforcement costs of the proposed undertaking.


87

IXInterbrew and Bass mergerThis chapter focuses on the case of the acquisition by Interbrew SA (Interbrew) of thebrewing activities of Bass PLC (Bass) for £2.3 billion. In July 2000 the proposedacquisition was referred to the UK competition authorities by the European Commission.In August 2001, after the UK Competition Commission (CC) completed its analysis andnecessary undertakings were fulfilled, the merger was consummated.

1 Background1

THE COMPANIES

Interbrew is the Belgian brewer of Stella Artois. Formed in 1987 it is the world’s secondlargest brewer and owns brands such as Boddingtons, Dos Equis, Jupiler, Labatt, RollingRock, and Stella Artois. In the UK market, Interbrew brewed and distributed Stella Artoisunder a licence it had with Whitbread. It then acquired the brewing business ofWhitbread PLC in May 2000.2

Bass Brewers was founded in 1777 in Burton upon Trent. Bass holds the UK’s oldestregistered trademark (1876). From 1967-1995 Bass was the UK’s largest brewer. It ownsthe Bass, Hooch, Caffrey’s, Carling, Grolsch (in UK), Tennents, and Worthington brands.

The CC estimated that the merger would create the largest brewer in Great Britain, witha market share of between 33 and 38 per cent and a portfolio of leading beer brands.3

The merger would have also brought together many of the UK’s best known and bestselling lagers and bitters (Stella Artois, Boddingtons, Worthington and Caffrey’s). Table 1provides information on the sales and concentration in the British beer market from 1996through 1999. Table 2 provides similar information on the on-trade beer distributionmarket in the UK.

1 The information in this chapter relies upon the report issued by The Competition Commission on the Interbrew SA and Bass PLCtransaction, “A report on the acquisition by Interbrew SA of the brewing interests of Bass PLC”, January 2001.

2 Report, Section 1.2.

3 Ibid., 1.3. The market share in wholesaling and distribution of beer was 33%.


88

TABLE 1

CONCENTRATION IN THE BRITISH BEER MARKET

Competition Commission data AC Nielsen data

1996 1997 1998 1999 1996 1997 1998 1999

Scottish 27.8% 26.7% 25.6% 26.0% 26.3% 25.8% 25.3% 25.6%Courage

Bass 21.9 23 22 23.4 21.1 21.2 21.4 21.8Brewers

Whitbread 13.5 13.8 14.2 14.5 13.8 14.5 14.7 15.2

Carlsberg- 15.4 14.6 13.2 11.8 13.2 12.5 12.1 11.7Tetley

Guinness 6 5.7 5.7 5.6 6 5.6 5.7 5.5

Anheuser- 2.5 3 3.4 3.2 1.6 2 2.3 2.2Busch

Wolverhampton 1.3 1.7 2.8 3.1 0.9 1.6 2.2 2.5Dudley

Greene King - 0.6 0.6 0.9 0.8 0.7 1.2 1.1

Others 11.5 10.9 12.6 11.5 16.3 16.1 15.1 14.3

HHI 1,720 1,691 1,566 1,625 1,542 1,519 1,503 1,545

Notional merger analysis including Heineken

Interbrew 36.8% 36.2% 37.8% 34.9% 35.6% 36.1% 37.0%

HHI 2,325 2,190 2,301 2,124 2,132 2,131 2,209

Change in HHI 634 624 676 582 613 628 664

Notional merger analysis excluding Heineken

Interbrew 31.6% 31.5% 33.2% 30.2% 30.9% 31.5% 32.1%

Heineken 5.2 4.6 4.6 4.7 4.8 4.6 4.9

HHI 1,997 1,898 1,996 1,841 1,836 1,844 1,894

Change in HHI 306 332 371 299 317 341 349

Note: “HHI” is the Herfindahl-Hirschman Index. The HHI is a measure of the concentration in a market and is obtained by summing thesquared value of the market shares of the firms in the market.

Source: Competition Commission 2001, Interbrew SA and Bass PLC.


89

TABLE 2

CONCENTRATION IN UK ON-TRADE BEER DISTRIBUTION

Market distributor 1996 1997 1998 1999

S&N 25.6% 25.8% 26.4% 26.4%

Tradeteam 22.1 22.2 23.3

WBC 10.5 10.2 9.5 9.4

Carlsberg-Tetley 12 12.4 11.4 12

Wolverhampton 2.3 2.9 4.6 5.2

Greene 1.5 1.6 2.1

Independent 6.3 6.2 5.9

Others 18.9 18.1 15.7

HHI 1,419 1,436 1,504

Notional post merger analysis

Interbrew 32.3% 32.7% 32.7%

HHI 1,869 1,857 1,944

Change in HHI 449 421 439

Source: CC based on questionnaire responses by brewers.

2 Competition commission analysis

MARKET DEFINITION

The CC defined three markets: (a) the supply of beer by brewers; (b) wholesaling anddistribution; and (c) the supply of beer by retailers. The report focuses on (a) and (b)which are the two markets that would have been affected by the merger. The geographicmarket for beer was defined as Great Britain.4 An additional distinction was made, at thewholesale and distribution level, between on-trade sales (pubs, hotels and clubs) and off-trade sales (retailers) given the difference on packaging and distribution requirements.5

At the time of the merger the consumption of beer had been declining steadily. This decline was forecast to continue in the future. Within this overall declining trend,there are some countervailing trends. For example, the consumption of lager hasincreased. Also, off-trade sales have grown while on-trade sales have declined.

4 Ibid. 2.26.The commission excluded Northern Ireland, which was considered a separate market from Great Britain and was notaffected by the merger (Interbrew had negligible sales in Northern Ireland).

5 Ibid., 2.27 Draught beer accounts for over 90 per cent of on-trade sales, while off-trade sales are mainly of packaged beer, distributedby the multiple retailers or by independent wholesalers and cash-and-carry operators.


90

BARRIERS TO ENTRY

The report concluded that entry was difficult because of scale economies in brewing and packaging. The CC determined that the substantial investments needed to build an efficient-scale brewery represented a barrier to entry. Also, the CC believed that theneed to invest in advertising and promotional expenditure represented an additionalbarrier to entry. These expenditures on advertising and promotion are sunk costs of entryas they would not be recoverable if entry failed.

With respect to distribution, an entrant was thought to have easy access to retailers in the off-trade segment.6 The CC concluded that access to on-trade distribution was abarrier to entry and expansion in brewing.7 Vertical integration (by brewers into pubownership) was not considered a major deterrent to entry because of the Beer Orders.8

Following the passage of the Beer Orders, the extent of vertical integration by brewersfell, especially among national brewers.9

PRICE DISCRIMINATION

Interbrew charged different prices according to the type of customer with whom it wasdealing. Interbrew offered more attractive prices to multiple retailers than it did toindependent free trade or independent wholesalers. The CC thought that the pricedifferences observed were not entirely explained by differences in costs of serving thesedifferent customers. As a result of this price discrimination, the independent free trade,who were generally charged a higher price, found it difficult to compete with the retailersthat had multiple sites and generally were charged lower prices. The CC believed thatthe merger would enhance the ability to price discriminate.

CONCLUSIONS

The CC concluded that the merger would strengthen Interbrew’s portfolio of leading brandsand lead to the creation of a duopoly - between Interbrew and Scottish & Newcastle plc(S&N).10 The CC was concerned that the duopolists may find it mutually convenient toraise operating margins and, by colluding, to raise net wholesale prices. The CC believedthat collusive behaviour would have been easier once the merger was implemented.

In its conclusions, the CC stated that consumers generally had quite inelastic demand for beer. Consequently, the CC believed that retailers would be able to pass higherwholesale prices that would result from the transaction through to consumers. The CCalso expected the additional revenues from the higher prices to lead to an increase innon-price competition. This non-price competition was likely to raise entry barriers. The promotion of the leading brands by the two major brewers would create strongerbrands which would impede the entry of new competitors into the market.

6 Ibid., 2.41.

7 Ibid., 2.49 and 4.43.

8 On 10 July 1989 the Government announced a series of measures designed to remedy the public interest detriments identified by theMMC 1989 report. Appendix 3.8.

9 Ibid., 2.43.

10 Ibid., 1.5.


91

The report also discussed the likely efficiencies arising from the merger. It concludedthat the merger11 would generate a market structure where Interbrew and S&N wouldhave the largest and most efficient distribution networks. The CC recognised thatcompeting brewers could rely on the centralised wholesaling and distributionarrangements offered by some multiple retailers. Even with such centralised services,entrants or smaller brewers would not have access to retailers that do not offer suchservices or the independent free trade. Thus, for these customers, Interbrew and S&Nwould effectively control the route to market.

Interbrew told the CC that it expected synergy benefits and cost savings from themerger. Interbrew also assessed the size of the benefits and the costs of implementingthese changes.12 The company believed that the synergies would enable it to continue to offer competitive pricing to all channels without further erosion of already tightmargins. The CC, however, believed that the sum Interbrew paid for Bass Brewers(above its own estimate of fair value) was a premium that limited “any beneficial effectsthat consumers would experience from those synergies even if Interbrew realised all the synergies it claims are available.”13

REMEDIES

Eventually, an agreement between the UK authorities and Interbrew was reached. This agreement allowed Interbrew to divest the England and Wales business of BassBrewers Limited to Coors Brewing Company.14


In its final report, the CC focused on the effect of the horizontal combination of the twofirms. For these reasons, this case might appear to not be appropriate given the subjectof this study. We believe that there are several possible bundling and portfolio issuesthat could have arisen during the investigation. Thus we will go through the possibleissues that could have arisen and discuss why we believe at the end of the day theseissues were not troubling.

Interbrew reported that it believed Bass Brewers’ products and its own product range werecomplementary (in the UK, Interbrew did not own or have a license to produce any majorstandard lager brand, while Bass Brewers did not own any major premium lager brand).In addition to the complementarities of the UK products, there were complementaritiesin the international operations. Bass operated in countries where Interbrew was notestablished-allowing easy international expansion for the Interbrew products. Similarly,Interbrew’s international reach would open up major new export opportunities for Bass

11 Ibid., 1.6.

12 The exact figures are omitted from the published version of the report for confidentiality reasons.

13 Ibid., 1.9

14 This divestiture was the brewing assets as well as the brands owned by Bass Brewers, which included Carling - the best selling lagerand beer brand in the U.K.


92

Brewers. Given the decline in demand for beer15 in the UK, the internationalopportunities would allow for the combined company to enjoy economies of scale.

There remain two possible areas of concern. First, the concentration in the beer distributionmarket could have effects in the beer market. Second, there is a question of whether thebroader portfolio of beer brands would allow the merged firm to exercise market power.

As discussed in Section 4.4.2 of the main report, gaining control of a complementarymarket such as distribution, can lead to entry deterrence. The entry deterrence arisesbecause of the need for new entrants to enter into two markets simultaneously. Theevidence in the beer distribution market, though, was that there were still opportunitiesfor new entrants to gain distribution of their products. Interbrew claimed that theindependent wholesale sector had increased in strength significantly in the two yearspreceding the merger. Two of the major independent national wholesalers were boughtby major drinks organisations (The Beer Seller by H P Bulmer and Matthew Clark byCanandaigua, then renamed Constellation Brands). Since neither of these distributorswas allied with either of the major UK brewers, the distributors should be willing todistribute a new entrant’s products (if such products warranted such distribution).

Also, both the CC and Interbrew found in logistic companies a viable means to accessdistribution. An example of such a company is Wincanton. S&N, Carlsberg-Tetley,regional brewers and independent wholesalers had logistics operations. Bass contractedout physical distribution to Tradeteam, a joint venture with Exel, which was managedindependently by Exel. While we do not have information on cost comparisons betweenin house distribution and logistic operators, given that existing companies use theselogistics operators, we believe that they are a viable alternative for distribution of beer.

Another question the merger raises is whether after the merger Interbrew would be ableto offer bundled discounts across its multiple different brands to the detriment ofcompetition. The bundle of different beer brands could be thought of as a variety bundle(these issues are discussed at length in Section 4.4.4.1 of the main report). In particular,after the merger, if there had been no brands sold off, Interbrew would have had anumber of different types of beer it could offer to its customers. By offering a bundledprice for its beers to a particular pub, Interbrew could have affected the demand for acompetitor’s brand of beer at that pub. The effect of this bundle would be similar to thatdescribed in Chapter VII, the case study of the Aspen case. There is an importantdifference, though, in that while there likely are consumers of skiing that are locked in to using the Aspen areas, it is not obvious why there would be a similar lock-in having todo with the brands of beer that are on sale at a particular pub. While the variety bundleof Interbrew beers would affect the demand for a competitor’s brand at particularestablishments, there does not seem to be any effect from this across a correctlydefined market (say all pubs in a particular area). If Interbrew was able to get all of the

15 The decline was from 65.5 million hectolitres in 1990 to less than 59 million hectolitres in 1999.Consumption decreased particularly inthe on-trade, reflecting the continued trend towards increased off-trade consumption driven in part by increased on-trade retail pricesand decreasing off-trade prices (which suggest that consumers are price sensitive).


93

pubs in a locality to accept its bundled offer, there could be issues, but given the currentmarket shares of brewers given in Table 1, such a concentration of sales seems unlikely.In other words, the harm from the variety bundles tends to be that they result in otherfirms and products being excluded from a market and we do not believe that Interbrewwould have the ability to cause such exclusion given its brands in the beer market.

There are some interesting empirical questions that can be raised in this market thatwould assist in evaluating whether this type of transaction would ever lead tocompetitive harm because of bundling of multiple beer brands. For example, when abrewer or distributor starts to offer a broader range of products, do competing brewersof one (or few) brands suffer? Also, it would be interesting to try and quantify thenumber of significant beer brands that would be necessary to exclude competitors. But, at the end of the day we reach a not very exciting conclusion - that there are notlikely to be any competitive harm from Interbrew’s ability to bundle more products.


94

XSMG SRH - ScottishRadio caseIn this chapter, we look at bundling of advertising across several different types ofmedia. We show that the structure of the media markets in the UK is such thatanticompetitive bundling is unlikely to occur. We conduct this analysis through theexamination of the recent acquisition of additional share ownership of SRH by SMG. In March 2001 SMG plc (SMG), a company with business interests in several mediamarkets, had increased its ownership of Scottish Radio Holdings plc (SRH) from around15% to 29.5%. This acquisition was evaluated by the UK Office of Fair Trading (OFT).1

After investigating the transaction, the OFT decided not to refer it to the CompetitionCommission and the transaction was consummated.

1 Background

THE COMPANIES

SMG is a media company with interests in television, radio, press, magazines, outdoor,and cinema advertising. SMG originated from the ITV franchise, Scottish Television,which acquired Caledonian Publishing (The Herald, a daily paper in Scotland, The SundayHerald, and the Evening Times) in 1996, Grampian Television in 1997, Pearl & Dean(cinema advertising) and Primesight (outdoor advertising) in 1999 and Ginger MediaGroup (including Virgin Radio and Virgin FM) in 2000. SMG also has a minority interest inthe national ITV breakfast-time broadcaster, GMTV (25% share), and Heart of Midlothianplc, the Scottish Premier League football club. In the year ending 31 December 2000,SMG had turnover of £300.5 million, pre-tax profit of £59.0 million and gross assets of£631.3 million.

SRH is the principal commercial radio operator in Scotland. The origins of SRH trace back to the assignment of a third commercial radio license in the UK in 1970. A group oflocal businessmen created a consortium in Glasgow and obtained the license. In 1974,they created Clyde radio. SRH is now running 9 local radio stations in Scotland and 43local newspaper titles primarily in Scotland and Ireland. In addition, SRH is active inoutdoor advertising. In the year ending 30 September 2000, SRH had turnover of £71.7million and pre-tax profit of £16.5 million. The analogue radio business profits accountedfor £11.1 million, while the press sector had more than £5 million profits.

1 “Completed acquisition by SMG plc of 29.5% shareholding of Scottish Radio Holdings plc,” Report under section 125(4) of the FairTrading Act 1973 of the Director General’s advice, 21 June 2001, to the Secretary of State for Trade and Industry under section 76 ofthe Act.


95

At the time of the investigation, SMG had no director on the board of SRH and was notinvolved in the day-to-day running of SRH’s business. SMG was the largest shareholderin SRH (the next largest held 10%), which gave SMG veto power on resolutions wheresuper majority voting was required. The OFT concluded that since SMG had the ability toinfluence the policies of SRH, the acquisition should be considered a qualifying merger.

The UK Broadcasting Acts of 1990 and 1996 limits cross-media ownership. Theserestrictions prohibit a company that controls an ITV license from also controlling radiolicenses in the same coverage areas. Thus SMG would not be allowed to increase itsownership in SRH above 30% under these broadcasting acts.

2 OFT analysis

The OFT reviewed the likely effect of the transaction in three markets. First, in radioadvertising, there was no overlap at the local level. SRH broadcasted solely in Scotland,and SMG broadcasted in London and nationally (Virgin Radio). After the transaction, theparties would hold approximately 14.8% of the revenue in the national radio advertisingmarket. Second, local newspaper advertising was a separate relevant market. SRHoperated 12 titles in Scotland, which were all weekly local papers, whereas SMG’s titleswere regional, or Glasgow area papers. Thus there was no overlap between the twocompanies and there was no competitive concern in this market. Finally, the market foroutdoor advertising was considered national, and the incremental increase in marketshare due to the merger was negligible. At the regional level the parties had shareslower than 25% in the sub-markets for 6 sheet and 48 sheet sites. The parties wouldhold less than 30% of the market for 48 sheets (20’ by 10’ billboards) and 96 sheets (40’ by 10’ billboards). Thus it was concluded that even after the transaction the outdooradvertising market was reasonably competitive with four credible competitors. Thus, theOFT concluded that the merger did not raise competition concerns in any of the threemarkets. The inquiry then analysed the portfolio effects of the merger.

Possible portfolio effects were considered given that the merger put together the leadingtelevision and radio stations in central Scotland, covering Glasgow and Edinburgh, aswell as SMG’s three newspaper titles. The aim of the analysis was to evaluate whetherthe merger would have reduced competition between radio and newspaper or televisionadvertising. OFT also evaluated the possibility that parties would have bundledadvertising packages across all its media interests to the detriment of customers.

The OFT was sceptical that economies of scale or scope could arise given the “disparateproduction, distribution and marketing processes for television, newspaper and radioadvertising.”2 For example, SMG reported to the OFT that the display media industrywas characterised by specialists “selling and buying of each particular media to reflectthe requirements of specialist advertising agencies.”3 There was also no evidence thatcustomers had a preference for purchasing all their advertising requirements from onesource. A SMG competitor that held businesses in radio, magazines and television, had

2 OFT report at page 3.

3 Ibid., page 3.


96

previously attempted a central selling strategy but then reverted to separate sellingarrangements. A third party, Carlton Sales Limited, sold the majority of its televisionadvertising revenues. Some evidence also was provided by SMG that in spite of beingactive across several media sectors, it never chose to integrate its sales division. Finally, the ITV television licence prevented conditional selling of advertising.

Given these characteristics of the industry, the OFT concluded that it was unlikely thatthe merger would result in anticompetitive effects from bundling different products. The OFT also concluded that the competitiveness of each individual advertising displaymarket, made refusal to supply an implausible strategy.


This case illustrates our theoretical point (described in section 7.1 of the main report)that firms can have multiple types of relationships at the same time. Advertising acrossdifferent mediums is complementary for some advertisers, while it is consideredsubstitutes for others. For example, a national advertising campaign is likely to view thedifferent mediums to be complementary. But, local advertising in papers and on theradio, is likely view the different mediums as substitutes. Taking an industry-wide viewwill tend to neutralise these two effects. We think it would be a mistake to look at a zeroaggregate effect and conclude the markets are independent. A better approach is to lookat how each of these groups are impacted separately. For the customers who see thedifferent mediums as substitutes, this transaction should be analysed as a horizontalmerger. For the customers who see the different mediums as complements, this mergershould be viewed with an eye towards potential bundling effects.

For the question of bundling effects, we examine the likelihood that there will be theability to exercise portfolio power from the transaction. These types of effects can arisebecause there are economies of scale and scope that can be enjoyed by the firm orthrough the artificial creation of such effects through the use of bundling and/or tying.There was no evidence that there were genuine economies of scale or scope that could beenjoyed across the different types of media advertising, so we will explore the possibilityof creating such economies through pricing and/or bundling of the various products.

The evidence suggested that in the past the customers had not welcomed efforts to bundlethe different types of advertising. We do not know if this failure in combining the salesof the different types of advertising was because of poor performance on the part of thefirm or if customers truly were not interested in bundling across the different advertisingmarkets. We would expect there to be a number of customers that advertise in all of thedifferent media.4 It would seem that these customers would welcome the opportunity toplace multiple adverts through a single seller. If so, this would be a case where bundlingcreates value, much as the way a car maker creates value by bundling together thesteering wheel with tyres. The failure of the other firms to construct a bundle of productssuggest that the consumer preference for a bundle, if it does exist, is not strong.

4 This commonality of customers across the different advertising media would be for both firms that are advertising and for theagencies that are creating and placing the advertisements.


97

A key fact to consider is that the OFT determined that there was competition in each ofthe individual markets. Thus our theory of bundling suggests that there is no opportunityfor bundled pricing to cause immediate competitive harm. As explained in Section 3.2 ofthe main report, in these competitive conditions, there is no way to create market powerthrough bundling. There is no Cournot effect with competitive markets, nor is there apotential to use bundling as an entry deterrent. Since all components are available atcompetitive prices, metering is also ruled out. Nor is there a view that a combined SMG-SRH would have a unique ability to offer a bundle. The one possible concern is theelimination of a competitive complements market, as in the ISO market with Kodak (asdescribed in Chapter VI above), but this issue is not relevant to the SMG-SRH case.

So, as the OFT concluded, each of the media markets is individually competitive, sothere is neither a horizontal effect to be concerned with nor a bundling effect to beconcerned with.


98

XIForeign packageholidays and insuranceThe media, especially when holiday times draw near, devote remarkable attention toholiday prices and practices in the travel industry. Not infrequently we read alarmingtitles such as “Anger Over Package Holiday Extras”1 or “The Tourist Trap in Your HighStreet”.2 These articles document anecdotal consumer outrages about pricing of holidaypackages. These outrages include excessive pricing for travel insurance and other extra,add-on prices that are not clearly advertised. At the beginning of 2002, two-thirds of thetravellers abroad bought their holiday insurance from travel agents. Our interest in thecase arises from the wide-spread strategy of bundled pricing of holiday packages in theUnited Kingdom. In this chapter we review the industry and the recent investigation intothis industry in the UK and show that the bundling has not lead to anticompetitiveeffects. Instead, the bundling of holidays and insurance appears to allow agencies andtour operators to hide the individual prices of the components of the bundle. The MMCpublished a report in 1997 which focused on the foreign package holiday industry. In thischapter we discuss the findings and conclusion of this report especially as they relate tothe tying of insurance to foreign holiday packages.

1 Background3

Following complaints by independent participants in the travel industry regarding increasingvertical integration in the market and possible anti-competitive practices, the UK Monopoliesand Mergers Commission (MMC) investigated the supply of travel agent services andtour operator services in relation to foreign package holidays in the UK. These wereholiday packages that involve transport between the UK and a foreign destination alongwith accommodation. The MMC concluded that there was some limited competitionbetween foreign package holidays and other types of holiday packages, but found thatthe independent holidays and domestic packages did not provide an effective constrainton prices for foreign package holidays. The MMC also concluded that while there wassome substitution between face-to-face sales and other sales techniques such astelephone and direct mail, the market was limited to face-to-face travel agency services.

1 “Anger Over Package Holiday Extras,” The Independent, Thursday 11 July 2002.

2 “The Tourist Trap in Your High Street”, Guardian Ltd, Saturday, 26 January 2002.

3 “A Report on the supply in the UK of tour operators’ services and travel agents’ services in relation to foreign package holidays,”United Kingdom Monopolies and Mergers Commission, December 1997.


99

Most foreign holidays from the UK are to destinations within continental Europe. The majorityinvolve air travel. Holidays are packaged by tour operators and sold to the public directly(approximately 80%) or through travel agents (20%). Table 1 provides market shares for1997 of the tour operators. Table 2 provides the same information for travel agents.

The Commission found that the tour operators engaged in practices which distortedcompetition and that their combined market share was over 25%. It concluded that therewas a complex monopoly situation in existence.4 The Commission found that the travelagents engaged in practices which restricted competition and that they held a combinedmarket share in excess of 25%. The Commission found, among other practices5, that thetying of travel insurance to discounted holidays was detrimental to the public interest.6

2 MMC report

With respect to tying, the Commission found that the parties engaged in the followingharmful practices. First, the MMC found that the tour operators were:

4 “Monopoly situations” can exist when there is at least a 25% market share by one member, according to section 7(1)(a) or (b) of theFair Trading Act 1973.

5 The other two relevant findings were “most favoured customer” clauses specifying that the travel agent offer the same discounts ona tour operators’ package holidays that it does on other operators’ packages and failure to sufficiently notify customers of ownershiplinks between tour operators and travel agents where vertical integration takes place.

6 MMC report at 47.

TABLE 1

Operator Parent Company Market Share

Thomson Tour Operations Ltd. Thomson Travel Group (TTG) 24.6%

Airtours Airtours 15.9%

First Choice Holidays & Flights First Choice Holidays 10.1%

Sunworld Ltd. & Thomas Cook Thomas Cook Group 4.2%

Inspirations East Ltd. Inspirations plc 2.3%

Other 42.9%

Total 100.0%

TABLE 2

Agent Parent Company Market Share

Lunn Poly Ltd. Thomson Travel Group (TTG) 23%

Going Places Leisure Travel Airtours 16%

Thomas Cook Thomas Cook Group 12%

A T Mays Carlson 6%

Other 43%

Total 100%


100

• pricing foreign package holidays so that the component parts of the total price,including any holiday insurance, were not transparent to the consumer;

• tying travel insurance to the purchase of foreign package holidays in certaincircumstances, particularly where a discount was offered.

Furthermore, the MMC found that the travel agents were:

• marketing foreign package holidays in such a way that the prices of the individualcomponents, including any holiday insurance, were not transparent;

• making the purchase by consumers of travel agent’s own travel insurance a conditionof obtaining a discount on certain foreign package holidays.

In theory, either tour operators or travel agents could have practiced tied discounts to thepurchase of travel insurance. The investigation found that for tour operators, “the practiceof tying discounts to the purchase of insurance occurs in only a small minority of all salesof foreign package holidays” and that it “does not constitute a practice which prevents,restricts or distorts competition.”7 However this was not the case for travel agents.Roughly 60% of all foreign package holidays sold by travel agents were discounted, and50% of those were conditional on the purchase of insurance. According to the MMC,the practice “allows the travel agent to inflate the advertised discount on the foreignpackage holiday by reason of the sometimes large margins made on the sale of insurance.”8

The travel agencies claimed that the practice benefited consumers by making it morelikely that they would be covered by insurance, tailoring the insurance to the specificholiday package, and offering overall cheaper prices. In what follows we describe the sales policy (with respect to bundling holidays and insurance) of the main factors in the travel industry and our view on the Commission’s investigation.

Thomson Travel Group (TTG)

TTG is the parent company to the tour operator Thomson Tour Operations, and the travelagent company Lunn Poly. Thomson Tour Operations did not offer any insurance-linkeddiscounts. Lunn Poly, however, “made its discounts conditional on the purchase of itsinsurance.”9 Since 1994, over 70% of Lunn Poly’s sales were packages that includeddiscounts and insurance. TTG argued that this strategy was a pro-competitive practice as it was an important driver of price competition. In addition, TTG argued that bundlingwas needed to “protect its own brand image from the adverse publicity which wouldfollow from one of its customers experiencing serious trouble overseas.”10

Airtours

Airtours runs both the Airtours tour operating company and the travel agency GoingPlaces Leisure Travel. An Airtours holiday requirement was that the customers take aninsurance policy of equal or greater protection than those offered by Airtours

7 MMC report at 40.

8 Ibid. at 40.

9 Ibid. at 132.

10 Ibid. at 177.


101

themselves. Its tour operating company offered discounts linked to insurance, whichaccounted for less than 5% of its sales. Going Places frequently offered such tied deals.Discounts were always conditional on the purchase of Going Places’ insurance.11 Itsview was that this bundling was not harmful to consumers as they are able to choosethe source for their insurance. Those consumers that are highly price sensitive inpurchasing foreign package holidays for their insurance, would seek information aboutinsurance options. Although Airtours considered a remedy to be unnecessary, theystated that any remedy should apply to “all tour operators and travel agents who wouldor might otherwise engage in the practice.”12

Thomas Cook Group (TCG)

Sunworld Ltd. and Thomas Cook Holidays (TCH) are tour-operating companies owned by TCG. TCG is also the parent company of the Thomas Cook travel agency. Sunworldand TCH have similar policies whereby holiday prices are not linked to the purchase of insurance, but adequate insurance is required. Less than 1% of their customerspurchase travel insurance from the TCG operators. However, the travel agent arm ofThomas Cook did make its discounts conditional on the purchase of insurance. TCGclaimed that this tie was clearly advertised. TCG believed that the discounts were notagainst the public interest since consumers were aware of the choices in insurance andwere “able to make price comparisons.”13 Similar to the other operators, TCG felt thatbundling facilitated product differentiation and enhanced competition and that consumersbenefited from the tie. TCG believed that by prohibiting the tying of insurance todiscounts, the result would be a decline in innovation, higher prices, and more uninsuredholidaymakers. TCG recommended the alternative remedy of “the introduction of astatutory cooling-off period of seven days for all sales of travel insurance, whether linkedto discounts or not...[to] give consumers time to seek alternative suppliers of insuranceand to compare their merits.”14

First Choice

Finally, First Choice tour operators had a policy requiring customers to have insurance(from either First Choice or any other source). For holidays purchased through agents,First Choice never tied its prices and discounts to insurance policies. Within the FirstChoice direct sales business (Eclipse), some deals were conditional on the purchase ofFirst Choice insurance. A small portion of the holidays were sold directly to consumers.According to First Choice, most travel agents required holidaymakers to purchase theagents’ insurance to receive discounts. First Choice noted that in the market forinsurance that should have been competitive, the prices for travel insurance productswere “remarkably similar” (only a 5p difference between insurance for 14 nights offeredby Lunn Poly, A T Mays, and Thomas Cook).15 First Choice believed that customers were

11 Ibid. at 185 and 186.

12 Ibid. at 191.

13 Ibid. at 194 and 195.

14 Ibid. at 199.

15 Ibid. at 201 and 202.


102

losing the benefit of part of the discount by buying overpriced insurance. It stated thatthe higher premiums were used to fund discounts and customers were discouraged bythis apparent discount from shopping around for alternative products and better value.First Choice stated that in its direct sales business it had “no alternative but to tie thesale of insurance to its discounts as its own headline prices had to be comparable withthose (already tied to insurance purchase) on the high street.”16 The company supportedthe remedy that prohibited the tying. It felt that this would increase consumerawareness of insurance sales and pricing. First Choice said that it would not favour aremedy that discouraged travellers from carrying insurance.

Independent participants in the travel industry argued that the practice of tying insuranceto discounts misled consumers and that the insurance prices offered in the tyingarrangement were higher than those available separately from other sources.

The MMC concluded that the practice of tying insurance distorted competition betweentravel agents. Lunn Poly, Going Places, Thomas Cook Group, A T Mays and other travelagents were found to have tied insurance. The MMC stated that besidesmisrepresenting prices, tying mislead consumers and was not in the public interest.

The MMC noted that the tying should be “distinguished from bundling various services,for example free taxi transfers from airport to hotel... Such services are sometimesreferred to as being offered ‘free.’ [The Commission did] not regard bundling asinherently objectionable as it does not mislead the consumer. An offer of ‘free’ travelinsurance does not mislead consumers as to the price of either the holiday or theinsurance element.”17

REMEDIES

The MMC recommended that insurance tying be prohibited. To prevent the tie just beingshifted upstream to the tour operator level of the transaction once the remedy is enforced,the MMC proposed applying the ban to both travel agents and tour operators, and to allcompanies regardless of size. Offers of “free” insurance would continue to be permitted.

On 10 August 1998, the Minister for Competition and Consumer Affairs signed an Orderunder the Fair Trading Act 1973. The Order made it illegal for a travel agent or touroperator to “discriminate against any customer in the price they charge for a foreignpackage holiday, or to make an additional charge, if the customer does not buy insurancewith the holiday.” The Minister opined “consumers should not be forced to take outinsurance which may not be competitively priced or does not meet their needs in orderto obtain a discount on a holiday. Nor should they have to pay more for their holidaysbecause travel agents are discouraged by tour operators from offering discounts theywould otherwise be prepared to offer.”18

16 Ibid..

17 Ibid., at 47.

18 DTI press release, 10 August 1998.


103

3 Current market conditions

Some four years have passed since the remedy was enacted. Below we describe how the major players have responded to these changes in the market.

TTG/Lunn Poly19

Lunn Poly now offers its foreign holiday packages with the option of purchasing non-TTGinsurance. Lunn Poly offers “silver” and “gold” insurance packages, as well as anexcess waiver and the chance for customers to supply their own insurance. It requirescustomers to carry some form of travel insurance. Lunn Poly requires travellers toprovide the company and policy number of the insurance company if insurance is notpurchased with the trip. In this case, the customer is also required to release Lunn Polyfrom any responsibilities or fault arising from inadequate insurance coverage. Lunn Polyhas chosen to offer other bundled discounts with its packages, such as free car upgrade,25% off per person, free theme park ticket upgrade, and £20 off when booking online.

Airtours/Going Places20

Going Places (now known as My Travel) now accepts insurance other than its own if the insurance gives comparable coverage to the holiday traveller as the insurance offeredby Going Places. Examples of ways in which My Travel now offers bundled discountsincludes resort transfers, pre-bookable seating, luggage allowance, free and reducedchild prices, free admission to airport play areas and children’s clubs at resorts, and £20off when booking online.

First Choice21

First Choice lists a total price with and without insurance for its online package holidays.Like other travel companies, First Choice requires the traveller to carry insurance.Examples of ways in which First Choice offers bundled discounts includes savings onairport parking, commission-free holiday money, free welcome pack, and £40 off percouple when booking online.

TCG22

The Thomas Cook Group offers “Classic Travel Insurance” and “Travel Insurance”packages to customers. Travellers can also opt to purchase insurance from otherproviders. Examples of ways in which TCG now offers bundled discounts include freeinsurance, up to 50% off, and free child places.

19 http://www.lunnpoly.com/etail/.

20 http://www.going-places.co.uk/.

21 http://www.firstchoice.co.uk/.

22 http://www.thomascook.co.uk/bold/homepage.htm?CC=1


104

Other

The Association of British Insurers (ABI) have a published information sheet that estimatesthe cost of holiday travel insurance at around £25-30 per adult for up to 15 days in Europeand the Mediterranean. For the rest of the world, comparable trip insurance would beabout £40-50, according to the ABI. Because of the high risks associated with ski vacations,winter sports holiday insurance is about twice as expensive as that for summer travel.Travel insurance can be purchased from a variety of sources, including the Post Office,American Express, Tesco, Boots, ASDA, and others. The results for similar packages for2 people travelling from the UK to Europe for a 14-day holiday are shown in Table 3.

TABLE 3

PRICE COMPARISON (2 PERSON 14-DAY EUROPEAN HOLIDAY)

Company Type Parent Basic Extended

Company Insurance Insurance

Price Price

Lunn Poly Ltd. Agent/Operator TTG £63.98 "Gold" £83.98

Thomas Cook Agent/Operator TCG £59.98 "Classic" £79.98

Travelcare Agent/Operator Co-operative Group £54.00

My Travel Agent/Operator Airtours £45.98 "Gold" £65.96

Direct Line Insurance £45.60

First Choice Operator First Choice £40.00

Trailfinders Insurance £39.00

Voyager Insurance £33.81 "Gold" £47.54

TPU Insurance £27.73

Worldwide Insurance £24.70 "Super" £30.40Travel Insurance

Primary Direct Insurance £24.47

CostOut Insurance £23.10 "Super" £24.20

Insure & Go Insurance £21.80 "Standard" £28.50


105

4 Economic analysis

Despite the remedy imposed by the MMC four years ago, travellers from the UK maystill be paying more for the convenience of purchasing insurance from travel agents (see Table 3). The BBC reported that as of early 2001, holidaymakers “could be payingup to four times” the lowest price for insurance. Up to 60% of the price of the insurancewent to travel agents’ commission.23

Similarly, The Guardian24 refers to a report by the Research Department that showed thatsingle trip coverage for a family of four in Europe could be obtained for just over £18 viaTravelPlan Direct and over £95 through travel agents’ Airtours Winter Sun (the ResearchDepartment rated the coverage supplied by Airtours as inferior to Travel Plan’s four starcoverage). Other newspapers and travel websites report similar figures with price ratiosfor the same products of 589%25 and 614%26 between direct insurance providers andtravel agencies.

BUNDLING AS A WAY TO OBSCURE PRICES

The bundling of insurance with foreign package holidays did not seem to lead to anycompetitive distortions in either the travel agent market or the tour operator market.Both of these markets seemed to be competitive both before the remedy was imposedas well as after. This is not a surprising conclusion since, as is shown in section 3.2 ofthe main report, there is no ability to create market power from bundling when themarkets that are bundled are competitive. There also appears to be no way to eliminatethe complementary industry that currently supplies travel insurance through this bundle,so this case is different from the Kodak case presented in chapter VI above. In mostcases, consumers were able to buy holiday packages and insurance separately. The maineconomic rationale for the operators and agencies to sell bundles of holidays andinsurance seemed to be to advertise holiday packages at very low prices, while stillearning a competitive margin through inflated insurance prices.

The tour operators generally felt that the bundles were an aspect of competition in theirindustry. Also, consumers appear to prefer to purchase bundles as they have continuedto purchase insurance from the tour operators and travel agents even though suchinsurance appears to be higher priced. We do not know whether the travel agents and/ortour operators informally discourage consumers from shopping around for insurance,which could explain some of the purchase activity by consumers of the higher pricedinsurance. There are also transaction costs for the consumer associated with obtaining

23 Http://news.bbc.co.uk/hi/english/uk/newsid_1130000/1130574.stm, “Clients Overpaying for Travel Insurance.” BBC News online, 22January 2001. Viewed 20 May 2002.

24 “Choose the Right Travel Policy”, Guardian Ltd, Wednesday 15 August 2001.

25 In 2001, insurance for a family of four to spend two weeks on holiday in Florida cost £46 from a direct insurance provider, while it cost£270 from Thomas Cook travel agents. “Family Travel, Everything on Holidays and Travel with Kids.” Family Travel online:http://www.family-travel.co.uk/practicalitites.

26 In July 2000 a family of four going to Spain could get higher levels of protection from a £22 policy purchased online from CostOut thanunder the £135 insurance sold through Thomas Cook. Wright, Simon. “Money: Don’t be Taken for a Ride on Travel Cover.” SundayMirror, 30 July 2000.


106

insurance from a third-party - both the cost of making the additional purchase as well asthe cost of having to prove to the tour operator that an adequate third-party insurancepolicy has been obtained. We do not know the relative importance of these two effects;whichever one is responsible, this does not appear to be a competition issue as thereremains competition in all segments of this industry.

The particular remedy that was imposed appears to be the correct one in addressing theproblems associated with the deceptive pricing. Forcing the tour operators and travelagents to not make discounts on part of the package contingent on purchasing the entirepackage allows consumers to value all of the components of the package separately.Figure 1 shows the costs of insurance from various sources since the remedy wasimposed. It might be worthwhile to investigate why consumers appear to still bepurchasing the majority of insurance from the high price source. If it turns out that thisbehaviour is the result of the added costs associated with buying the insurance from a different source than the travel company, then perhaps there should be an effort to tryand reduce these costs. If the current behaviour is the result of some other type ofactivity that is distorting consumer purchasing decisions (such as heavy handed saletactics by the travel agents), then that perhaps should be looked at to see if there issomething that can be done to allow consumers to make the lower cost purchases.

FIGURE 1

PRICE COMPARISON (1 PERSON 14 DAY EUROPEAN HOLIDAY)

£0.00

£5.00

£10.00

£15.00

£20.00

£25.00

£30.00

£35.00

£40.00

£45.00

My Travel (Going Places)

Thomas Cook

A T Mays

Lunn Poly Ltd.

Abbey National

Post Office

Bradford & Bingley

Columbus Direct

19972002


107

XIIBT telephone andinternet bundlingIn November 2000, British Telecom (BT) announced a new tariff package (SurfTime) forresidential customers, including two new pricing schemes for off-peak internet accesson an unmetered basis. One year earlier, in December 1999, BT had announced an initialSurfTime version of the SurfTime product, which was investigated by Oftel. The DirectorGeneral of Telecommunications investigated this original pricing offer. The investigationexamined whether the new pricing scheme would leverage BT’s dominance in themarkets for local and national voice calls into the internet access market. “Specificallythe Surf element would be offered in a bundle together with voice calls at a marginalprice below cost.”1 In this chapter, we review this case, especially in light of the resultsin section 4.4.1 of the main report.

1 Background

In November 2000, BT announced BT Surf Together. The product provided low voice call prices and unmetered off-peak internet access calls. BT Talk & Surf Togetherincluded the low voice rates of BT Together, plus unmetered off-peak fees for bothinternet access calls and local voice calls.

BT’s Surf product does not provide internet access on its own, rather it allows internetaccess calls to be placed during off-peak times at an unmetered rate. The user mustseparately contract with an Internet Service Provider (ISP) to receive Internet service.Off-peak periods were Monday to Friday 6pm through 8am and all weekend. UnmeteredInternet calls are those calls placed to particular numbers so the call traffic can besegregated from the switched network traffic. This allows BT to separate internet trafficfrom voice calls. By keeping the Internet traffic separate, BT is able to more efficientlyoperate its network. BT has offered the stand-alone product, SurfTime, since June 2000.2

Oftel investigated the initial SurfTime product, announced in December 1999. Their viewwas that it was “important that other industry players should be able to compete withSurfTime.”3 Oftel’s supervision caused BT to alter its plans for SurfTime, to allow othertelecom companies’ access to wholesale services in order to compete.4

1 Investigation By The Director General Of Telecommunications Into The BT Surf Together And BT Talk & Surf Together PricingPackages, 4 May 2001, ¶4.

2 Ibid., ¶5.

3 Oftel letter to Internet Protocol Operators dated 23 February 2000.

4 BT revised the price of SurfTime and it reduced the original five tariffs down to three.


108

BACKGROUND ON INTERNET CALLS

One of the primary methods for consumers to access the Internet is through the telephonenetwork. In the UK, prior to 1998, consumers paid both local call charges and an ISP’smonthly fees in order to access the internet. The ISP provider Freeserve was launched in1998 by the Dixon group. Freeserve was the “first company to offer a mass market,subscription-free Internet access service for the cost of a local phone call.”5 In other words,Freeserve did not charge its customers for providing the ISP part of the Internet connection.Freeserve was able to make such an offer to its customers because of the structure ofTelecom regulation. In particular, for any phone call that is placed, the carrier that terminatesthis call is paid a per minute fee by the firm that originates the call. So, in this example, aBT customer who subscribes to Freeserve pays BT the cost of a local phone call to connectto the Internet but pays nothing to Freeserve. Since this call is terminated on the Freeservenetwork, BT pays Freeserve the terminating access rate. The level of the terminatingaccess was set by the regulators at a level such that BT paid “more than two-thirds of itscall revenue to Energis, the telco whose ISP division, Planet Online, [operated] Freeserveon behalf of Dixons. Given sufficient economies of scale, that income alone is enough tofund the service.”6 The SurfTime product by BT was designed in such a way that itwould not have to pay such a high rate for its terminating access to ISPs. Therefore, thisproduct was a threat to the business model upon which Freeserve was based.

In June of 2000 (at the time BT launched SurfTime), Freeserve initiated a programwhereby members would subscribe to BT SurfTime for £5.99 and get £1.00 a monthback from Freeserve. Now owned by the Wanadoo Group, Freeserve continues to offera pay-as-you go free service (for the cost of local calling), as well as £10.99 unmeterednight/weekend (with BT Surf), and £13.99 unmetered anytime service.7 In addition toFreeserve, there are other providers of these services. The offers from some of thesecompetitors include as well as the BT offerings were:

• Tiscali offered free Internet calls during non-peak hours for £11.99 per month;

• One.Tel offered a flat rate of £12.99 per month for unlimited Internet use; and

• The offers from BT and its subsidiaries shown in Table 1:8

2 Oftel’s analysis

The EC has determined that dominance can be presumed if a firm has over 50% of themarket. In the UK, “BT holds a dominant position in the market for wholesale callorigination on fixed telecommunications networks.”9 Oftel also believed that BT was“dominant in the markets for local and national retail voice calls for residential customerson fixed telecommunications networks.”10

5 Freeserve Company Information, www.freeserve.com. Viewed 30 May 2002.

6 Phil Wainewright. “Free access reaches new users.” ASPnews.com 20 February 1999.www.aspnews.com/news/article/0,,4191_373921,00.html Viewed 30 May 2002.

7 Freeserve Company Information, www.freeserve.com. Viewed 30 May 2002. “Freeserve Now Pays Its Members to Surf the Web.”Freeserve Press Release 01 June 2000.

8 The ISPs BT Internet and BT Connect are wholly owned subsidiaries of BT, making BT a vertically integrated supplier in internetaccess. BT provides call origination, termination, internet access calls, and internet service.

9 Oftel investigation, ¶9.

10 Ibid., ¶9.


109

In the report, Oftel stated that it was concerned of the potential for anticompetitivebehaviour through the horizontal leveraging of a firm’s market power in one retail marketinto another market. “In this case, the existence of a bundle containing both retail voicecalls and retail internet access would not, on its own, be anti-competitive, because BTalso offers tariff packages which do not bundle these services together... But... the issuein this case is the price at which the Surf element is being offered in the two Packageswhich were investigated.”11

11 Ibid., ¶11.

TABLE 1

INTERNET ACCESS AND ISP PRODUCT PRICES 21 MAY 2002

Product Price Description

BT SurfTime Anytime £19.99 unlimited on-peak and off-peak surf calls

BT SurfTime Evenings £5.99 unlimited off-peak surf calls& Weekends

BT Together £11.50 special call rates

BT Together with unlimited £14.50 unlimited local off-peak calls, special rates local calls for other calls

BT Together with unlimited £18.50 unlimited UK off-peak calls, special rates for UK calls other calls

BT Together with unlimited £14.50 unlimited surf calls (0844 04) off-peak, special Surf calls rates for other calls

BT Together with unlimited £19.50 (see combination of local and surf above)Local & Surf calls

BT Together with unlimited £23.50 (see combination of UK and surf above)UK & Surf calls

BT pay-as-you-go calls £0.00 1p per minute day, 0.6p per minute evening, to ISPs and 0.5p per minute weekend

Freeserve No Ties £0.00 customer pays phone account charges, but no fees from Freeserve

Freeserve AnyTime £13.99 unlimited internet calls on-peak and off-peak

Freeserve HomeTime £10.99 unlimited off-peak internet calls (£5 to BT for SurfTime and £5.99 to Freeserve)

Freeserve Unlimited Time £10.00 service suspended 08/01. £10 in national & international calls or unlimited internet anytime.

BTopenworld Anytime £15.99 unlimited internet calls on-peak and off-peak, no additional call access charges

BTopenworld Pay-as-you-go £0.00 calls charged at BT's standard local call rate

BTopenworld for BT Surf £6.99 unlimited off-peak internet access, standard localpackages call rate on-peak. BT SurfTime required (min £5)

BTopenworld Broadband £29.99 unlimited internet access via broadband connection (installation fee additional)

Sources: www.bt.com, www.freeserve.com


110

A second possible source of anticompetitive behaviour under investigation was verticalleveraging. Oftel reviewed whether BT could distort competition in wholesaletermination of internet calls and retail internet access. BT had “the potential to restrict ordistort competition in retail internet access markets and wholesale call termination ofinternet calls through the control of the margin available to competitors between theprices for its retail unmetered tariffs and the charges for wholesale call origination.”12

Internet Access Markets

The UK national retail market for Internet access is comprised of the provision of ISPservices and Internet access calls from the consumer to the ISP. Oftel concluded that theretail market includes both of these components because in order for a consumer to usea minute of Internet access, the user needs both a minute of access calls and ISP services.

A second internet access market is that of wholesale services. Wholesale servicesconsist of call origination (from consumer to the point of hand-over) and Internet calltermination (from hand-over to the ISP). Both halves of the call are necessary for anInternet access call to exist. Therefore, the report concludes that there are separatewholesale markets for call origination and call termination.

Wholesale Call Origination Market

The report examines the wholesale call origination market. Oftel concludes that mobilenetwork origination is not an effective demand side substitute for fixed line originationand so should not be included in this market. It was estimated that BT’s share ofwholesale call origination by volume was 81% for residential customers and 73% acrossall customers, the Director concluded that BT was dominant in the market for wholesalecall origination on fixed networks in the UK.

Retail Inland Voice Call Market

In its investigation, Oftel focused on inbound calls on fixed networks. BT’s share in themarkets for local and national retail voice calls was 73-77%. In addition, BT’s high ratesof return as well as barriers to customer service switching led the Director to concludethat BT was dominant in the markets for local and national retail voice calls by residentialcustomers on fixed networks in the UK.

Pricing Analysis

In order to assess whether the SurfTime product was being offered below cost in BT’spackages, Oftel proposed to compare prices and costs and “assess whether an equallyefficient competitor... could at least match BT’s price for Surf.”13

Oftel considered whether the marginal price of Surf (within the packages) was higherthan long run incremental costs (LRIC) of providing this service. The marginal price for

12 Ibid., ¶12.

13 Ibid., ¶52.


111

Surf in the BT Surf Together package was £5.40 per month, and in the BT Talk & SurfTogether package £5.00. The price of SurfTime Evenings and Weekends was £5.99, andSurfTime Anytime was £19.99 (see Table 2).

The report concluded that retail costs are sensitive to assumptions on underlying driversof the costs and that “the information relating to retail costs... was not sufficient to supporta conclusion that Surf, at the current prices, is being offered below cost in the Packages.”14

OFTEL’S CONCLUSIONS

Oftel looked at the potential impact of BT’s pricing behaviour on retail internet accessand wholesale call termination markets. Surf, as an off-peak metered package, facescompetition from 24/7 metered packages. The price differential between the 24/7 andoff-peak packages is generally not very big (as low as £3-4 per month).15 Even if “BT’sSurfTime products were to become the primary or sole, sustainable off-peak unmeteredpackage, competition with 24/7 unmetered packages would remain. Therefore, theDirector considers it unlikely that the Packages, as currently priced, will have a materialanti-competitive effect in the relevant internet access markets.”16

14 Ibid., ¶60.

15 The off-peak packages are not competitive if the consumer is likely to make many calls to an ISP during prime hours.

16 Ibid., ¶61-5.

TABLE 2

HISTORICAL BT PRICES 4 MAY 2001

Product "Price

(per month)"

BT Surf Together £14.99

BT Together £11.99

difference £3.00

value of inclusive call allowance £2.40

Marginal price for Surf on BT Surf Together £5.40

BT Talk & Surf Together £19.99

BT Talk Together £14.99

Marginal price for Surf on BT Talk & Surf Together £5.00

BT SurfTime Evenings and Weekends £5.99

BT SurfTime Anytime £19.99

Source: Oftel report.


112

With respect to the cross subsidisation of internet provision in the bundle, the reportstated that “it is not clear that, at the current prices, BT is or will price Surf below cost inthe Packages. Furthermore, the Director considers it unlikely that, even if the price werebelow cost, a material anti-competitive effect would result on current prices, because ofthe availability of sustainable 24/7 packages.”17

Therefore, Oftel decided not to pursue the investigation, finding that BT’s packages ofbundled products did not constitute an infringement of the Competition Act 1998.

3 Economic analysis

The basic concern that triggered the Oftel enquiry was the idea that by bundling unmeteredInternet calls provision to telephony services BT could expand its market share in theformer in an uncompetitive fashion. One might be concerned about BT using its profitsin the market in which it has market power to finance losses in the other market. Suchbehaviour, on its face, is not rational. As we have shown in section 3.2 of the main report,such cross-subsidy on its own can not accomplish the creation of market power. Theredo not appear to be the characteristics in this industry to allow for the creation of marketpower. We will look closer at whether the bundling will allow BT to create entry barriers.

Often a firm with market power may try to preserve that power by reducing the competitivepressure of an existing (one-product) rival. The bundling can make a difference by makingthe BT bundle preferred to the other, single-product firms’ products. If these advantageslead to the single-product firms exiting the industry and/or new firms being unable toenter both industries at the same time, then there can be long-run anticompetitive effects.If, on the other hand, the bundling leads to other firms creating a bundle of products,then often the outcome is stronger competition than if there is just competition of singleproducts. We now examine the evolution of the ISP industry in the aftermath of theintroduction of the SurfTime packages to try and see which outcome appears more likely.

Since early 2000, several companies have been competing with BT by offering phone serviceand Internet access. For example, One.Tel provides unmetered internet access18,Telewest19 announced its unlimited package on 14 February 2000, and NTL20 has beeneffectively providing unmetered access since April 17, 2000. In addition, Tiscali, offers itscustomers a bundle of telephony and unmetered Internet access. This entry of firmsproviding bundles suggests that there is tough competition in the market.

There also has been entry by single-product firms, as AOL launched its first unmeteredinternet connection (with no charge the internet phone calls) which is on sale now at£15.99. This entry suggests that the bundle is not necessary to compete. That AOL hashad limited success with this offering in the UK, though, suggests that the bundle isactually important.

17 Ibid., ¶66.

18 “One.Tel Customers Go Surfing!” 02/03/00, http://www.onetel.co.uk/xhtml/ie/residential/ index.php?Pg=about&story=2, viewed June 2002.

19 “Forget BT’s SurfTime, Telewest launches SurfUnlimited”, 14/2/2000, http://news.zdnet.co.uk/ story/0,,t269-s2076985,00.html, viewed June 2002.

20 “The war is on: NTL offers free access”. Tuesday 7/3/2000, http://news.zdnet.co.uk/ story/0,,t269-s2077512,00.html, viewed June 2002.


113

Overall, the market for ISP services and the bundle of telephony and ISP has notbecome dominated by BT. The “number of new dial-up access packages (unmetered orpay-as-you-go), indicates a highly competitive market in which the consumer benefitsfrom a wide choice of prices and packages.”21 Table 3 provides information on theevolution of the ISP market over the last several years. This Table is consistent withsignificant competition and new entry and is not consistent with BT being able todominate the market with its bundled offering.

This evidence suggests that indeed the Oftel Director was correct in allowing BT tolaunch the bundle package. Since the introduction of this bundle by BT, Freeserve hascontinued to be the most popular residential ISP over the last 12 months (18%) andAOL’s market share has grown considerably from 8% to 17% (May 2000 - 2001). BT Internet and more recently, NTL, also enjoy significant market presence (8-12%) andthere are many smaller ISPs representing 28% of the market who, at under 2% each.Thus there is no evidence that BT’s bundled offering has given it an advantage over itsrivals. Furthermore, it is possible that the offering of this bundled product has led togreater competition than if BT had not started offering the bundled product.

TABLE 3

ESTIMATE OF INTERNET SERVICE PROVIDER SUBSCRIBER SHARE

May 00 Aug 00 Jan 01 Feb 01 May 01

Freeserve 29% 27% 21% 21% 18%

AOL 8% 10% 15% 13% 17%

BT Internet 5% 10% 10% 12% 12%

NTL 4% 5% 9% 8% 8%

LineOne 3% 2% 2% 3% 4%

BT Click 2% 5% 2% 3% 3%

Virgin Net 5% 3% 5% 3% 3%

Supanet 3% 3% 2% 2% 3%

Netscape 3% 3% - 2% 3%

Tiny Online 2% 3% 2% 2% 2%

Others 34% 29% 32% 31% 28%

Source: Oftel quarterly residential survey.

21 Oftel’s 2000/01 effective competition review of dial-up Internet access, 30 July 2001, ¶3.62.


XIIIMobile telephonesIn this chapter we examine the pricing strategies in the mobile telephony market. This is an industry that has a history of many different types of bundled offerings -bundling of minutes (e.g. contracts with a fee for a given number of minutes of calls),bundling of complementary services (e.g. SMS and voice), and bundling of complementaryproducts (handsets and service contracts) to name a few. We look at a number of thesebundles and evaluate whether there is likely an anticompetitive effect from this pricingbehaviour. We begin by describing the structure of the mobile telephony industry in the UK. To assist in the understanding of this structure, we also provide a briefbackground on some of the technology involved in the industry. We also contrast thestructure of the market in the UK with other countries in Europe.

1 Background

Currently there are four facilities-based mobile operators in the UK: Vodafone, O2

(formerly known as BT Cellnet), Orange, and T-Mobile (formerly known as One to One).In addition, there is a major Mobile Virtual Network Operator, (“MVNO”) - Virgin.1 Also,there is a fifth potential operator, Hutchinson 3G, that has purchased a license and isplanning on starting to offer 3G mobile services in the UK shortly. Table 1 providesinformation on the market shares of the four facilities-based mobile operators over thelast five years. As can be seen, using share of subscribers as the measure, while Orangeand T-Mobile started small, by 2001 the four operators are all similar in size.

TABLE 1

OPERATOR SHARES: MOBILE SUBSCRIBERS

Vodafone O2

T-Mobile Orange

Dec-01* 25% 25% 23% 27%

March 01 28% 26% 21% 25%

March 00 32% 27% 18% 22%

March 99 37% 30% 15% 17%

March 98 38% 34% 13% 15%

March 97 40% 38% 9% 13%

*The methodologies for calculating subscriber numbers changed between March and December 01. These changes may understateVodafone market share in particular. See Oftel’s Mobile Market Review statement in September 2001, ¶2.36

114

1 Virgin Mobile is a 50:50 joint venture between Virgin and T-Mobile. Virgin supplies the handsets and bills customers while it uses the T- Mobile network to provide the service.


115

Using revenues as an alternative measure, the four operators are not so similar in size.These shares are presented in Table 2. In particular, Orange and T-Mobile havesignificantly smaller shares when measured by revenue as opposed to subscribers.

TABLE 2

ESTIMATED RETAIL REVENUE BY NETWORK OPERATOR

Vodafone Cellnet One2One Orange

1997-98 44% 33% 10% 12%

1998-99 40% 34% 12% 14%

1999-00 36% 36% 13% 16%

Qrt 2 00-01 35% 29 15% 21%

Notes: last row is based on the first two quarters of 2001-02 only.

Sourcel: Oftel Market Information and Oftel Estimates.

These operators offer a wide range of calling plans to customers, including manydifferent pre-paid plans as well as contracts with a fixed fee being paid for a certainnumber of minutes each month. The current customer mix between these two types of plans is 70% on pre-paid and 30% on a contract. Overall, 72% of the population in the United Kingdom owns a mobile phone. In 1996, this penetration was only 11%, socustomer growth has been strong over the last six years. The contract calling plan iscommon in the United States, but not in many other markets. For example, in the Italianmarket, which is the market with highest mobile penetration in Europe, there aregenerally no such contracts offered by the operators. We use this alternative form ofcompetition to evaluate the competitive effect of this form of bundled prices.

Table 3 provides information on the growth of mobile subscribers in the UK. Thesegrowth figures are shown for overall subscribers as well as broken down into pre-paidand contract subscribers. As can be seen, the growth in the pre-paid subscribers hasbeen particularly strong in the last several years.

TABLE 3

GROWTH OF THE UK MOBILE SECTOR

Mar-97 Mar-98 Mar-99 Mar-00 Mar-01 Dec-01

Post pay 7,109 8,556 10,439 12,164 13,745 13,650

Pre pay - 467 4,438 15,021 29,707 31,243

Total 7,109 9,023 14,878 27,185 43,452 44,893

Source: Oftel publication available at: http://www.oftel.gov.uk/publications/mobile/ctm_2002/mobile_market.pdf .


116

Another feature of mobile competition in the UK is the subsidisation of handsets. In theUK, most of these handsets, are made to work with a single operator’s network, i.e. theyare “locked.” By locking the handset, it is not possible for the handset to be used withother operators’ SIM2 cards. Recently it has become possible to purchase SIM cardsseparately from handsets, and by paying a fee of up to £20, the handset can beunlocked.3 The policy of having a handset tied to a particular network was not followedin Italy, Portugal, and Finland. We use this contrasting form of competition to explore theeffects of this particular bundle.

2 Analysis of pricing plans

THE RELEVANT MARKET

Oftel is currently investigating termination charges in the UK mobile industry. In previousreviews of the mobile sector, Oftel defined a single relevant market for the provision ofall mobile telephony services to end-users. This definition was based on the belief thatthe choice consumers face is the selection of a particular mobile operator to purchasethe entire bundle of mobile services. This approach focused on the possible constraintson a firm’s ability to set prices and earn profits. These constraints arise primarily on thedemand side and stem from the consumers’ ability to switch to alternative products inresponse to price variations. Oftel concluded that fixed services were not part of thismarket because there is not sufficient substitution between fixed services and mobileservices. In its current investigation, Oftel has concluded that there is a separate marketfor call termination on each mobile network. Oftel believes that each mobile operator hasmarket power in the supply of mobile termination to its own network due to the lack ofcompetitive pressures. In addition, though, Oftel seems to believe that call origination iscompetitive and that all of the mobile operators compete in a single market for this product.

PRICING PLANS

There is an extensive array of price offers available to consumers in the UK. In principlehaving such a large number of different price plans can have one of two effects. First, itmay allow a firm to price discriminate among various different types of customers.Second, it may lead to each consumer getting a price and product that best suits his orher needs. If the market for call origination is competitive, then the mobile operatorshave no market power so they can not price discriminate. So we would presume thatconsumers are benefiting from having many different choices. Normally we wouldbelieve that such an array of choices (without market power) would increase consumerwelfare, but it appears that this might not be the case. In particular, it seems that theprice plans are difficult to compare and hence it seems unlikely that most consumers arepurchasing the best plan for their particular calling patterns.

2 Mobile phone handsets contain a ‘subscriber identity module’ (SIM) which is used to identify which network the handset is connectedto, for billing and service access purposes. The SIM card is basically a computer chip that contains the identity of the subscriber andthe detail of the calling plan.

3 While it is likely that there was handset unlocking in the past, it is only recently that this option has been openly advertised, see forexample T-Mobile site on pay as you go (PAYG) price plans, www.t-mobile.co.uk.


117

In the UK, each provider has a number of contracts available to consumers. Thesecontracts vary in the monthly subscription fee, the number of bundled minutes provided,the amount of free text messages provided, as well as other attributes. Additionaldifferences include standard call costs, off peak call costs, international calls cost, textmessages costs, one off connection fee, and voicemail retrieval costs. For example,Vodafone offers contracts with bundles of 20, 60, 200, 400, 800, and 1100 minutes,each of these can be combined with a given number of text messages for free (none, 50or 300).4 Unused minutes from month one are available in month two. If they are still notused up, they are carried forward to month three. If at the end of month three theminutes originating in month one remain unused they are subsequently lost. Outstandingtext messages are forfeited every month. Similarly, O2 offers bundles of 30, 50, 100, 200minutes, combined with internet access, and text message bundles (none or 25, 50,100, 300). T-mobile offers bundles of off peak minutes for free, or provides cheaper callsmobile-to-mobile (both within network, or on all mobiles depending on prices) whenpurchasing a bundle of peak minutes. Orange offers bundles that range from 50 minutesto 10,000 and that includes call cost computation by the second and daily bundles of free minutes. In addition to these individual packages, group options are also available.For example, Orange offers packages that can include from 2 to 4 people to share 200free minutes a month, with cheaper calls and SMS within the group.5

In addition to the different number of minutes in these packages, the mobile companieshave different policies with respect to any unused minutes from one billing period to thenext. The roll over of minutes allows consumers to smooth out idiosyncratic events thatcause excessive calling in certain months. Some of the plans allow minutes to roll overand others that do not allow any roll over. In addition the terms of how long the roll-overminutes last and the amount of minutes that can be carried vary.

With all of these variations across the different plans (both within a given network aswell as across the different networks), the comparison of prices is a difficult exercise.With such complexity, consumers probably focus on a subset of these features, say thecost of the bundle of minutes, and choose a plan accordingly. There are two problemswith such a mechanism for selecting plans. First, the other attributes of the plan canhave a significant impact on the overall cost. Second, in most cases consumers cannotcorrectly anticipate their use of the mobile phone, so the selection of a particular bundleof minutes often is not the optimal bundle for that consumer. In fact, given all of thevariables that comprise the various mobile contracts, it is highly unlikely that a consumercan ex-ante know the average cost of calls for any particular calling plan. This inability toevaluate the prices arise because of the complex set of products that are bundledtogether in the various calling plans. In principle, this non-transparency of prices does nothave anticompetitive consequences but it does lead to a decrease in consumer welfare.6

4 Monthly price plan connections are normally charged at £35. Peak rates refer to 8 am - 7 pm. Monday to Friday evening, off peak,refer to 7p.m - 8 a.m. Off-peak weekend rates refers to midnight Friday to midnight Sunday.

5 In addition to all of these contract price plans, each operator has a number of pay-as-you-go pricing plans.

6 The loss of consumer welfare is not coming from firms gaining surplus, so this is not the type of loss of consumer welfare that thecompetition laws generally are addressing.


118

While this is not a competition problem, there can be consumer benefits arising fromchanges in the market. The goal of this regulation should be to improve the informationconveyed to customers. While it is relatively easy to construct a calculator that woulddetermine the lowest cost plan for a given calling pattern, it is not possible to know theexact pattern of calls and hence the input to this calculator is at best a poor estimate ofactual calling patterns. To be able to correctly evaluate the cost to a consumer of aparticular plan, the consumer would need to know how many calls he or she would likelymake in a given month, the average duration of these calls, the network of the recipientof these calls, the time of day of the calls, and likely some other attributes. Therefore, the chance that any particular consumers, even if he or she expendssignificant time and energy investigating all of the options, has the cheapest plan is quiteunlikely.7 The inescapable conclusion is that perhaps the mobile telephony market is onewhere more options to consumers does not lead to higher consumer welfare.

But, once this conclusion is reached, the solution to this problem is not obvious. If thereis regulation that limits the number of plans that any operator can offer, then it becomeseasier for the mobile operators in any market to coordinate their activities. Also, while itis unlikely that any particular customer has his or her optimal plan, if there are fewerplans being offered, one result could be that the distance that a consumer is from his orher optimal plan is greater under the restricted number of plan options. Thus, a limitingof the number of plans offered is unlikely to help consumers and could well result inharm. Another idea would be to require more information to be available to consumerson the pricing of services. It appears, though, that most information that is relevant isalready available as to the cost of the plan to a consumer and the difficulty in evaluatingalternative plans is an inability to predict ones exact usage patterns. The only potentiallyrelevant price that is not readily available is the price charged to others when they call aparticular mobile user’s phone. We will return to this idea below when we discussterminating access.

Another way to make customers more able to take advantage of price differences isthrough the use of multiple SIM cards in mobile phones. In Italy, Finland and Portugal,SIM cards have never been locked to mobile phones. Hence, customers can freelyswitch SIM cards within a handset in order to take advantage of price plans that havedifferent time-of-day pricing. Until recently, there were two primary ways of doing thisswapping: using a mobile phone which has space for only one SIM card at a time, whichconsumers manually change; and using a mobile phone which has space for more thanone SIM card at a time, which consumers can then toggle between SIM cards withouthaving to manually change the SIM card. With this second method it is still necessary toswitch on and off the mobile phone each time the customer wants to switch SIM cards.Also, the size and weight of these dual SIM card mobile phones is greater than manyhandsets that are in use in the UK today.

7 In fact this has resulted in one mobile operator in the United States guaranteeing that if a consumer on an alternative carrier brings inhis or her bill, that this operator will show that it has a plan that would be cheaper than the consumer’s existing plan. It is likely(although not advertised by this operator) that the same could be said for its own customers.


119

The proportions of mobile customers using more than one SIM card in May 2001 were -13% of mobile customers in Italy (5.8 million users), 7% (272,000 users) in Finland, and4% (288,000 customers) in Portugal. On average, multiple SIM users were youngerindividuals and those without a fixed phone at home. Most of these customers switchtheir SIM cards manually, while phones which held more than one SIM card at a timewere less widely used - 27% in Italy, 18% in Portugal, and 16% in Finland. The choice ofwhich SIM card to use depended on whether a call was made to a fixed or a mobilephone, which mobile network they were calling, and whether they had freeminutes/credit to use. On average, consumers evaluated which SIM to use for about halfto three-quarters of their calls.

The proportion of mobile customers, who had previously used a different network thanthe one they currently use, was only slightly higher in Italy than in the UK. However, inItaly, switching networks, packages, and tariffs was significantly higher among multi thansingle SIM users. Therefore, we would expect that the introduction of dual-SIM card useinto the UK would be an easy way to arbitrage between different network offers. In amarket like the Italian one, demand for these products developed early (even when thecard swapping was entirely manual and cumbersome).8

A final proposal to try and increase consumer surplus in the mobile sector would be formobile operators to allow switching between its various plans by consumers. This wouldallow consumers to learn about their actual usage and then to adjust their calling plans to obtain the best price. Currently most mobile operators allow such costless switching if a consumer moves to a more expensive plan (i.e. a plan with a larger number ofincluded minutes), but not necessarily if the consumer wishes to use a smaller numberof minutes. The problem with this option is that the operators all subsidise the purchaseof the handset, and if a customer might be switching to a less profitable plan, then theoperator likely will provide less of a subsidy to the handset. The effect on the willingnessto subsidise handsets will increase if there is a requirement to allow consumers toswitch between different operators without incurring any switching costs. On balance,we are concerned about the lack of transparency of mobile pricing, but we are notconvinced that there is a solution available to improve the situation.

3 Analysis of terminating access9

An issue that is currently being investigated in the UK is whether the mobile operatorsare setting the price of terminating a call on their networks too high. This is the price thatone telecommunications carrier pays to another when a call is placed from the firstnetwork to the second. This is not a cost that is directly paid by consumers, but since itis a cost incurred by the mobile operators, the overall price level must be set at a levelthat allows the recovery of this cost (if the firm is to remain in business). There aretechnological features that make the termination of calls to mobile phones more costly

8 The article that announces the launch of GIANO starts as follows “it will not occur you anymore to see someone dismantling itsmobile to change SIM”, which indicates how the practice was common and “visible”,http://www.ilnuovo.it/nuovo/foglia/0,1007,81710,00.html.

9 CRA is currently assisting one of the mobile operators with the investigation of terminating access charges by mobile operators in the UK.


120

than to fixed lines,10 but these cost differences are not nearly as great as the pricedifferences between terminating on a fixed line vs. on a mobile line. In fact, the pricescharged for terminating access appear to be significantly higher than the incrementalcost of providing the terminating service.11

There are two issues that could be addressed as to these termination charges. First, whether these prices are in some sense excessive and hence causing harm toconsumers. Second, whether there are some distortion to competition being caused by these prices that may need addressing by the competition authority (even if theparticular prices do not directly harm consumers). It is the second of these that is thefocus of this section.12

A possible consequence of having the terminating access priced too high is the creationof apparent economies of scale (similar to the effect discussed in Chapter VII, the AspenSki case study and section 4.4.4 of the main report). This arises in the case of mobilecalls because a call that is terminated on a carrier’s own network does not result in atermination fee being paid to another carrier. This causes the mobile operators to offercheaper call rates when a mobile consumer calls another mobile consumer on the samenetwork as compared to calling a fixed line or a customer of another mobile network.13

This discount for within networks calls gives an incentive to create a network ofconsumers (e.g. families and businesses) to subscribe to a single mobile operator. Such a network, as it grows would appear to perhaps lead to a tipping of the market,leading inexorably to a single dominant network. But we do not believe that this wouldbe the outcome. The reason why is that there is no reason for firms to follow thisparticular pricing pattern, especially if on average mobile customers receive as manycalls as they place. With a balanced level of calls, a mobile operator can charge itscustomers the same price whether a call is placed to it’s own network or to a differentnetwork. The ‘lost’ revenue from having to pay a high termination rate to other mobileoperators is made up through calls placed from that other mobile operator onto its ownnetwork. Similarly, while it may appear that high termination rates would raise a barrier to new entry, such is not the case using this exact same logic.14 The one type oftelecommunications operator that would not be able to protect itself in the face of hightermination charges is a land-line operator.15 Thus the high mobile termination chargesseem to have no effect on competition but result in land-line consumers providing asubsidy to mobile customers through high mobile termination charges.

10 This is due to the fact that knowing the location of the call receiver helps to minimise the distance between the interconnection pointand the receiving caller. In fixed line calls, the number being called allows the originating carrier to identify the geographical position ofthe receiver and hence to terminate the call at the optimal location. With mobile calls, this geographic determination can not be doneas well.

11 Note that since mobile telecommunications operators incur significant fixed and sunk costs, only by charging above incremental cost isthe operator able to remain in business. Therefore, observing that prices are above incremental costs is not sufficient to determinethat the prices are anticompetitive (or excessive).

12 The investigation currently being conducted in the UK appears to be focussing on the first of these issues. There are many reasonswhy the prices being high might not cause direct consumer harm. For example, if the mobile operators price all of their services usingRamsey prices, then since likely terminating access is the service with the most inelastic demand, this is the service that will have thehighest mark up (in order to meet an overall zero profitability constraint).

13 T-Mobile has started offering a pricing plan that does not contain this differential pricing depending on the network of the mobile thatis called.

14 In fact, if the new entrant does offer the same price regardless of the network being called, this could be a marketing advantage that itcould use to gain more subscribers. If an acquaintance complains about having to pay more for calling a subscriber on the newnetwork, that person can be told that the new network allows subscribers to call anyone for the same rate and does not discriminate.

15 In theory there is no reason that a fixed-line operator could not achieve the same type of outcome as the mobile operators, but inreality the fixed-network operators are regulated and thus unable to set the same termination rates as the mobile operators.


121

Therefore, it does not appear that high termination charges will lead to a distortion ofcompetition. One might ask, though, whether competition will drive these terminationprices down. Pricing plans offered by most mobile operators have the cost of calls madeand the cost of received calls bundled together in the subscription rate. Thus, there is a non-transparency of prices. Further, even though the different mobile operators havesomewhat different termination rates, generally there is a single price for calling any of the other networks charged to subscribers. Mobile consumers in general care moreabout direct costs incurred to them rather than what others pay to call them, so it isdifficult for competition for new consumers to drive the price of termination down. But, if there is no distortion in competition, there is no reason that such a driving downof prices is necessary.

4 Conclusion

As is clear from above, the issue with mobile phones is the non-transparency of prices.New technology that allows consumers to implicitly unbundle these prices could be away to increase consumer welfare. The new generation of handsets that allow for quickand easy substitution of SIM cards will make it difficult for mobile operators to bundlecheap outbound calls with high termination fees. All a consumer needs to do is get twoSIM cards, one with low outbound and high inbound rates and the other with highoutbound and low inbound rates. Then, the consumer can receive calls on one SIM cardand place calls on the other. We believe that improving the transparency of price plans offers and the development of technologies that will allow handsets to operatesimultaneously on several networks should enhance consumer surplus.


122

XIVJefferson Parish HospitalThe Jefferson Parish Hospital (JPH) is located in Louisiana, in a suburb of New Orleans.JPH signed a contract in February 1971, just before it opened its East Jefferson Hospital,between the hospital and an anaesthesiologists group, Roux and Associates. Under thecontract, any anaesthesiologist designated by Roux would be given hospital privilegesand be a member of the hospital’s medical staff. Under the contract, the anaesthesiologiststhat made up Roux and Associates would only practice at JPH. Also, the hospital wouldallow no anaesthesiologist to practice at the hospital if he or she were not a member ofRoux and Associates.

The hospital would purchase all necessary drugs and other supplies for theanaesthesiologists. The fees for anaesthesiological services were billed separately to thepatients by the hospital and divided equally between Roux and the Hospital. After thiscontract was entered, a board-certified anaesthesiologist specialising in obstetricanaesthesiology, was denied privileges at JPH. This anaesthesiologist, Dr. Edwin G.Hyde, sued for a declaratory judgment that the contract was unlawful (by violatingSection 1 of the Sherman Act) and an injunction ordering JPH to appoint him to thehospital staff. This case study examines the facts surrounding the case and whether ornot the contract between JPH and Roux was the type of bundling agreement that wouldlead to anticompetitive effects.

1 Background

At the time of the trial, approximately 875 operations were performed each month at JPHin 13 separate operating rooms. The District Court ultimately found that there were noanticompetitive consequences from the contract and hence no violation of the ShermanAct. The Court of Appeal reversed this decision and found that the exclusive dealingcontract was per se illegal. The Supreme Court heard the case and ruled in March 1984that the contract was not per se illegal and that there was no violation of Section 1.

Given the nature of medical operations, any users of a hospital’s operating room mustalso purchase anaesthesiological services. Further, only physicians that are givenprivileges at a particular hospital can work at that hospital. Thus, anaesthesiologicalservices are consumed in fixed proportions with a hospital’s operating room services.1

1 This is a bit of an oversimplification as depending on the type of operation, the amount of time and attention required of theanaesthesiologist can vary. We return to this point below.


123

Further, it seems unlikely that there is market power in the supply of anaesthesiologicalservices as entry by new anaesthesiologists would be relatively easy (and there were no suggestions in the case materials that Roux had as members the majority of theanaesthesiologists in the New Orleans area). Thus on its face, the combination ofanaesthesiological services and hospital services appears to be the type of product thatwould fit the Chicago School model as set out in section 3.2 of the main report. Underthese conditions, tying leads to no anticompetitive consequences.

2 Case history

In July 1977, Dr. Hyde sued for a declaratory judgment, claiming that the contractbetween JPH and Roux was unlawful and to obtain an injunction ordering the hospital to appoint him to the hospital staff.2 Hyde’s complaint alleged violations of Sherman Act as well as state competition law.

The United States District Court for the Eastern District of Louisiana rejected theseclaims3, finding that the anticompetitive consequences of the contract were minimal and were outweighed by the benefits from the contract. The District Court found thatthe exclusive contract resulted in higher quality services being provided at the hospital.The District Court also suggested that the principles of per se liability might not apply to cases involving the medical profession.

The United States Court of Appeals for the Fifth Circuit ruled only on the Sherman Actclaim. The Court of Appeals rejected as clearly erroneous the District Court’s finding thatthe exclusive contract was justified by quality considerations4 since the exclusivecontract was not necessary to achieve these ends. The Court of Appeals reversed thisdecision, deciding that the contract was per se illegal.5 It held that the case involved a tying arrangement because the “users of the hospital’s operating rooms (the tyingproduct) are also compelled to purchase the hospital’s chosen anaesthesia service (thetied product).”6 The court defined the relevant geographic market for the tying productas the East Bank of Jefferson Parish and found that JPH possessed “sufficient marketpower in the tying market to coerce purchasers of the tied product.”7

Finally, in 1984, the United States Supreme Court reversed the Circuit Court.8

The Supreme Court found a wider definition of the geographic market and found that the hospital lacked sufficient market power in this market to apply a rule of per seillegality to the tying arrangement. It also found that the contract did not unreasonablyrestrain competition. The Supreme Court found that there was no increase in price nor a diminution of quality of anaesthesiological services from the exclusive contract.

2 The credentials committee and the medical staff executive committee recommended approval of Dr. Hyde’s application for medicalprivileges, it was the hospital board who denied the issuing of these privileges because of the contract with Roux and Associates.

3 The United States District Court for the Eastern District of Louisiana, 513 F Supp 532. 1981.

4 See 686 F.2d, at 292.

5 The United States Court of Appeals for the Fifth Circuit, 686 F2d 286. 1982.

6 Id., at 289.

7 Id., at 291.

8 The United States Supreme Court, 466 U.S. 2. March 27 1984.


124

3 Economic analysis

In this case, as in many, the court decision revolves around the market definition -whether the geographic market for JPH is small or large and hence whether JPH hassignificant market power or not. Not having access to the information as well as notbeing directly relevant to this document, we will not examine the market definition inquestion. We first assume that the Supreme Court correctly defined the market and that JPH does not have significant market power in this market. With this assumption,we analyse the likely effect of the exclusive contract and also evaluate the likely motivesfor the parties that entered into the contract.

The two separate markets that could be affected by this contract are the hospital marketand the anaesthesiology market. In terms of the hospital market, it was found that therewere a number of hospitals in the area and that there was effective competition in thismarket. So, up until the time of the lawsuit, the contract was found to have no effect onthis market. Furthermore, since anaesthesiology is a small part of the total basket ofservices offered by a hospital it is unlikely that the contract with Roux had an effect onthe entry conditions into the hospital market.9

In terms of the anaesthesiology market, the Supreme Court found no evidence thatthere had been a diminution of the quality in this market and that there was no effect onprices in this market. Consumers were still able to choose a different anaesthesiologist if they wanted, by having surgery performed at a different hospital.10 Furthermore, therewas no evidence that anaesthesiologists were being blocked from practicing in the area,just that they could not practice at JPH.11 If all of the hospitals in the relevant geographicmarket had exclusive deals with anaesthesiologists, then there could be entry barriersraised in the anaesthesiologist market. In such a situation, though, it is not clear why thehospitals would enter into the exclusive contracts because these would lead to marketpower in the anaesthesiology market - a complementary product to the hospital services.

Thus, we do not see any competitive harm from these exclusive contracts in the short-run. If over time all hospitals enter into exclusive deals, then there might be lesscompetition between hospitals and/or entry barriers into the anaesthesiology market. To try and determine if either of these effects was driving the parties to enter into thesecontracts, we try and determine what is the motivation of the two parties.

9 This conclusion is reinforced by the fact that, in general, states control the entry of a hospital into an area, so the greatest barrier toentry for a new hospital is not whether it has access to anaesthesiology services but getting a host of other regulatory approvals fromstate governments.

10 Surgeons generally have privileges at multiple hospitals, so if a surgeon and/or a patient wanted a particular anaesthesiologist whowas not a member of Roux & Associates, then this surgeon could perform the operation at a different hospital. There was no evidencethat this type of switching behaviour was not available to patients in the Jefferson Parish area.

11 There is no information on how difficult it was for an anaesthesiologist to join Roux and thus gain access to JPH.


125

WHY DOES THE HOSPITAL PREFER EXCLUSIVE CONTRACTS?

JPH entered into this contract because it saw the contract as a source of competitiveadvantage, as a way to achieve cost savings, or both. As outlined above, under theassumption that JPH does not have market power, there does not appear to be anadvantage (at least in the short run) to JPH from this contract. It was suggested by theSupreme Court that perhaps JPH saw the bills for the anaesthesiology services as a wayto obtain more money from health insurance companies, but this is not really acompetition issue and is a contracting issue between health insurers and JPH. In termsof cost savings, JPH could view the contract as a way to ensure a high quality ofanaesthesiologists without having the costs associated with having to monitor thesephysicians. Roux will be responsible for acquiring anaesthesiologists and then JPHaccepts these onto its medical staff. Not having to monitor the anaesthesiologists andincur the costs of granting privileges to these physicians might be a significant costsavings to JPH. A test of this as a hypothesis for a reason to enter into such a contract isif JPH has similar contracts with other groups of specialised physicians. We do not knowif there were other such contracts.

Finally, we note that if one believes the leveraging theory of tying then JPH, by enteringinto this contract, is creating market power for the anaesthesiologists. In such a situation,the contract could lead to higher prices for a complementary good (anaesthesiologyservices) and thus harm the hospital. Under the contract, it is not clear how the prices forthe anaesthesiological services are set, but since JPH will be billing for these services so it will be aware of the level of pricing even if it does not set these prices. This pricetransparency to JPH of the anaesthesiology prices should give it at least some controlover the level of these prices. Thus it seems that JPH has at least in part protected itselffrom being harmed by having a complementor charging prices that are too high.

WHY WOULD ROUX AND ASSOCIATES PREFER THE EXCLUSIVITY?

We also need to examine why Roux would prefer these exclusive contracts. On its face,it seems that the contract would allow Roux to have some amount of market power. As was pointed out by the court, patients and surgeons would have the ability to moveto a different hospital if an anaesthesiologist not belonging to Roux was the preferredphysician for a particular operation. Since Roux was not allowed to practice at otherhospitals under the terms of the contract with JPH, Roux would not be able to gainmarket power across hospitals. Thus, it appears unlikely that Roux would be able to gainsignificant market power from the agreement. It is possible that there are efficienciesassociated with the Roux group being the only group of anaesthesiologists that will beworking in the JPH operating rooms, so there can be a single set of practices andprocedures. Both JPH and Roux will be willing to work together to improve techniquesand practices without having to worry about any improvement being copied by otherhospitals or anaesthesiologists. In other words, the contract will allow for an alignmentof the hospital and the anaesthesiologists that could lead to quality and cost savings forboth parties.


126

Thus, on balance, it appears that there were efficiency reasons behind the signing of thecontract that resulted in no competitive harm from the exclusive contract. The exceptionto this conclusion would be if ultimately all of the hospitals in the area signedcomparable contracts to the JPH/Roux contract. This might ultimately lead to a lesscompetitive anaesthesiology market that could lead to harm to consumers. Theprotection that the market has against such an outcome, though, is that the hospitalsshould not be willing to be a party to such an outcome, so if there were danger fromsuch an outcome we would expect the hospitals to protect itself from the harm it wouldincur from having a complementary service priced high. On the other hand, if there wasmuch to be gained, we might expect to see integration of the anaesthesiologists and thehospitals, so then the hospitals can gain from any market power. We do not know howthe market in the New Orleans area has evolved over time, so we can not evaluatewhether this is the outcome in this market. But this case appears to support theconclusion in section 6.2 of the main report, that tying allows for a higher quality productand likely does not lead to anticompetitive effects.


127

XVReferencesAdams, William J. and Yellen, Janet L.1976. “Commodity Bundling and the Burden ofMonopoly.” Quarterly Journal of Economics, 90 (Aug.): 475-98

Bakos, Yannis and Brynjolfsson, Eric. 1999. “Bundling Information Goods: Pricing, Profits,and Efficiency,” Management Science.

Brandenburger, Adam and Barry Nalebuff, Co-opetition, Doubleday, 1996.

Case C-333/94 P Tetra Pak International SA v Commission of the European Communities14 November 1996 (appeal, Fifth Chamber).

Case T-83/91 Tetra Pak v Commission [1994] ECR II-755 (Court of First Instance, 6October 1994).

Commission Decision 92/163/EEC of 24 July 1991.

Cournot, Augustin. 1838. Recherches sur les principes mathematiques de la theorie desrichesses, Paris: Hachette. English translation: (N. Bacon, trans.), Research into theMathematical Principles of the Theory of Wealth (James and Gordon, Mountain Center:CA 1995).

EC competition policy newsletter, 1994. Volume 1. Autumn/Winter.

Hart, Oliver D. 1995. Firms, Contracts, and Financial Structure. Oxford University Press.

McAfee. R. Preston, McMillan, John, and Whinston, Michael D. 1989. “MultiproductMonopoly, Commodity Bundling, and Correlation of Values,” Quarterly Journal ofEconomics, 104 (May): 371-84.

Nalebuff and Lu (2001) “A Bundle of Trouble” Yale School of Management working paper.

Nalebuff, Barry (2000) “Competing Against Bundles” in Incentives, Organization, andPublic Economics, (P. Hammond and G. D. Myles, eds.) Oxford University Press: London.

OECD Competition Policy Roundtables, 1996. N.8 , OECD/GD(96)131.


128

Official Journal of the European Communities, 1992. No L 72/42, 18/3/1992.

Official Journal of the European Communities, 1997. No C 27/5, 25/1/1997.

Patterson, Donna and Carl Shapiro, “Transatlantic Divergence in GE/Honeywell: Causesand Lessons,” 16 Antitrust 18, 21 (Fall 2001)

Stefan Schmitz (2002) “How Dare They? European Merger Control and the EuropeanCommission’s Blocking of the General Electric/Honeywell Merger” presented at Can We Regulate Competition Internationally? A Case Study of the Attempted GE/HoneywellMerger Symposium at the University of Pennsylvania, April 12, 2002

Whinston, Michael D. 1990. “Tying, foreclosure and exclusion,” The American Economic Review.


Printed in the UK on recycled paper with a minimum HMSO score of 75February 2003 Department of Trade and Industry. http://www.dti.gov.uk

© Crown Copyright.DTI/Pub 6522/0.3k/02/03/NP URN 03/ 612


Bundling, Tying, and Portfolio Effects: Part 2 Case …€¦ · Bundling, Tying, and Portfolio...

Documents

Transcript of Bundling, Tying, and Portfolio Effects: Part 2 Case …€¦ · Bundling, Tying, and Portfolio...